Formalizing and achieving multiparty agreements via commitments

2025年研究生考试考研英语(二204)复习试卷与参考答案一、完型填空（10分）Part A: Cloze TestRead the following passage and choose the best word or phrase to fill in each of the blanks. Each blank has four choices marked A, B, C, and D. You should choose one answer and mark the corresponding letter on Answer Sheet 2.The rise of the Internet and social media has dramatically changed the way people communicate. (1) __________, these technological advancements have brought both benefits and challenges.1.A. HoweverB. FurthermoreC. NeverthelessD. ThereforeIn the past, communication was primarily (2)__________through letters and phone calls, which were time-consuming and limited in terms of (3) __________.2.A. conductedB. transmittedC. exchangedD. achieved3.A. speedB. reachC. clarityD. frequencyToday, (4)__________communication is instantaneous and allows for global connectivity. People can (5)__________with anyone, anywhere in the world, in just a few clicks.4.A. oralB. writtenC. digitalD. visual5.A. interactB. correspondC. correspond withD. communicateHowever, (6)__________these advantages, there are concerns about the quality of communication. The (7)__________of communication through social media can lead to misunderstandings and misinterpretations.6.A. DespiteB. In light ofC. ConsideringD. Given7.A. speedB. volumeC. diversityD. complexityFor instance, (8)__________language often lacks the nuances and subtleties that are present in face-to-face interactions, which can (9)__________to miscommunication.8.A. informalB. formalC. writtenD. spoken9.A. contributeB. resultC. leadD. deriveMoreover, the (10)__________of social media can also have negative impacts on mental health. Excessive use of social media can lead to (11)__________and feelings of isolation.10.A. convenienceB. popularityC. accessibilityD. prevalence11.A. anxietyB. depressionC. fatigueD. stressTo mitigate these negative effects, it is important for individuals to(12)__________their use of social media and focus on(13)__________communication.12.A. controlB. reduceC. manageD. limit13.A. digitalB. writtenC. verbalD. face-to-faceIn conclusion, while the Internet and social media have revolutionized communication, it is crucial to recognize both the benefits and the challenges they present. By being mindful of our communication habits and seeking a balance, we can harness the power of technology while protecting our mental well-being.14.A. HoweverB. FurthermoreC. NeverthelessD. Therefore15.A. conductedB. transmittedC. exchangedD. achieved16.A. speedB. reachC. clarityD. frequency17.A. oralB. writtenC. digitalD. visual18.A. interactB. correspondC. correspond withD. communicate19.A. DespiteB. In light ofC. ConsideringD. Given20.A. speedB. volumeC. diversityD. complexityAnswers:1.A2.C3.B4.C5.A6.A7.B8.A9.C10.D11.B12.C13.D14.A15.A16.B17.C18.A19.A20.B二、传统阅读理解（本部分有4大题，每大题10分，共40分）第一题Read the following passage and answer the questions that follow.The rise of e-commerce has transformed the way people shop, creating both opportunities and challenges for businesses. Online shopping has become increasingly popular due to its convenience, wide variety of products, and competitive pricing. However, this shift has also led to the closure of many brick-and-mortar stores and has raised concerns about the future of traditional retail.1、Why has online shopping become increasingly popular?A. It is less convenient than traditional shopping.B. It offers a wider variety of products.C. It is more expensive than traditional shopping.D. It is less competitive than traditional shopping.2、What is one of the main reasons for the closure of many brick-and-mortar stores?A. The rise of e-commerce.B. Increased competition from other businesses.C. Higher operating costs.D. Lack of customer interest.3、What concerns have been raised about the future of traditional retail?A. The decline in sales at physical stores.B. The potential loss of jobs in the retail sector.C. The reduction in customer satisfaction.D. The increase in the number of online scams.4、According to the passage, what is one of the advantages of online shopping?A. It requires customers to leave their homes.B. It offers limited customer service options.C. It can lead to a decrease in the variety of products.D. It is more time-consuming than traditional shopping.5、What is the author’s main point about the impact of e-commerce on traditional retail?A. E-commerce is solely beneficial to consumers.B. E-commerce is causing the demise of traditional retail.C. E-commerce and traditional retail are complementary to each other.D. The impact of e-commerce on traditional retail is minimal.Answers:1.B2.A3.B4.B5.B第二题Read the following passage carefully and answer the questions below.In the age of information, the way we consume and process information has undergone a dramatic transformation. The advent of the internet and digital technology has revolutionized the way we access knowledge, communicate, and learn. One of the most significant changes is the shift from traditional print media to digital media.1、The first paragraph of the passage introduces the topic of:A. The impact of digital technology on traditional media.B. The evolution of information consumption over time.C. The role of the internet in modern society.D. The challenges of digital literacy in the information age.2、According to the passage, which of the following statements best describes the transformation in information consumption?A. There has been a gradual shift from print media to digital media.B. There has been a complete elimination of print media.C. The consumption of both print and digital media has decreased.D. The popularity of print media has remained consistent.3、The author mentions “the advent of the internet and digital technology” as a significant factor. What does this imply about their impact?A. They have had a minimal impact on our lives.B. They have revolutionized the way we access and process information.C. They have only affected certain segments of the population.D. They have been detrimental to our ability to learn.4、The passage suggests that the shift to digital media has led to:A. An increase in the amount of time people spend reading.B. A decrease in the quality of information available.C. A more diverse range of information sources.D. A reliance on technology for all forms of learning.5、What is the overall tone of the passage?A. CriticalB. NeutralC. EnthusiasticD. PessimisticAnswers:1、B2、A3、B4、C5、B第三题Read the following passage and answer the questions that follow.In recent years, the rise of social media has dramatically changed the way we communicate and interact with each other. Platforms such as Facebook, Twitter, and Instagram have become an integral part of our daily lives, allowing us to connect with friends and family across the globe. However, this convenience has come at a cost, as social media has also been linked to various negative effects on mental health.1、The passage mentions several social media platforms. Which of the following is NOT mentioned?A. FacebookB. LinkedInC. TwitterD. Instagram2、According to the passage, what is the primary concern regarding social media’s impact on mental health?A. It increases productivity in the workplace.B. It enhances social connections.C. It has a negative impact on mental health.D. It improves communication skills.3、The author suggests that the convenience of social media is:A. the only benefit of using these platforms.B. outweighed by its negative effects.C. a minor aspect of social media use.D. the main reason for its widespread popularity.4、Which of the following is an example of a negative effect of social media on mental health mentioned in the passage?A. Improved job opportunities.B. Increased self-esteem.C. Higher levels of stress and anxiety.D. Enhanced creativity.5、The author’s tone towards social media can best be described as:A. enthusiastic and supportive.B. critical and concerned.C. neutral and objective.D. negative and dismissive.Answers:1、B2、C3、B4、C5、B第四题Reading Passage 1Questions 1-5 are based on the following passage.In the United States, the history of women’s education dates back to thecolonial period. During this time, most women were educated at home, with the help of their mothers and other family members. However, as the country grew and the demand for educated women increased, the need for formal education for women also grew. The first women’s college, Mount Holyoke Female Seminary, w as founded in 1837 by Mary Lyon. This college was a significant step in the history of women’s education, as it provided a place for women to receive a higher education.After the Civil War, the number of women’s colleges in the United States increased dramatically. Many of these colleges were founded by women who were educated themselves and believed that education was essential for women’s advancement. One of the most influential women’s colleges during this time was Vassar College, founded in 1861. Vassar was the first college in the United States to offer a co-educational curriculum.In the late 19th and early 20th centuries, the role of women in society began to change. As more women entered the workforce, the need for higher education became even more important. Women’s colleges began to offer more professional and vocational programs to prepare women for careers in medicine, law, and other fields. This period also saw the rise of the women’s suffrage movement, which advocated for women’s right to vote. The fight for suffrage brought women together and highlighted the importance of education in achieving equality.The 20th century was a time of significant change for women’s education. The number of women attending college increased dramatically, and the numberof women earning college degrees also grew. In 1972, Title IX of the Education Amendments was passed, which prohibited discrimination based on sex in any educational program or activity receiving federal financial assistance. This law had a profoun d impact on women’s education, as it opened the door for more women to participate in higher education and pursue their careers.Today, women’s education has become an integral part of American society. Women are attending college and earning degrees in all fields of study. The history of women’s education in the United States is a testament to the determination and resilience of women who have fought for the right to be educated.1、What was the main purpose of the Mount Holyoke Female Seminary?A、To educate men.B、To provide a place for women to receive a higher education.C、To train women for teaching.D、To offer vocational programs.2、What was the significance of Vassar College during the post-Civil War period?A、It was the first college to offer a co-educational curriculum.B、It was the first women’s college to offer professional and vocational programs.C、It was the first college to admit African American students.D、It was the first college to offer a degree in women’s studies.3、What impact did the women’s suffrage movement have on women’s education?A、It led to the creation of more women’s colleges.B、It highlighted the importance of education in achieving equality.C、It resulted in the passage of Title IX.D、It reduced the number of women attending college.4、How did Title IX of the Education Amendments affect women’s education?A、It increased the number of women attending college.B、It reduced the number of women attending college.C、It had no impact on women’s education.D、It increased the number of women earning college degrees.5、What is the main point of the passage?A、The history of women’s education in the United States is a testament to the determination and resilience of women.B、Women’s education has always been a prio rity in the United States.C、The United States has always had a high percentage of women attending college.D、The role of women in society has not changed over time.三、阅读理解新题型（10分）Reading Comprehension Part B (New Type)PassageIn the era of digital transformation, the role of data analytics indecision-making has become increasingly significant. Organizations across various sectors are leveraging data analytics to gain insights, predict trends, and improve their operations. However, with the exponential growth of data, the need for skilled professionals in data analytics has surged. This passage discusses the importance of data analytics in modern business and the skills required to excel in this field.QuestionRead the following passage and answer the questions that follow.PassageData analytics is the process of examining large sets of data to uncover meaningful patterns, trends, and insights. It involves various techniques, such as statistical analysis, data mining, and machine learning, to extract valuable information from raw data. In today’s business environment, data analytics plays a crucial role in several aspects:1.Strategic Decision-Making: Data analytics enables businesses to make informed decisions based on factual evidence rather than intuition or guesswork. By analyzing historical data, companies can identify trends and patterns that may not be apparent through traditional analysis methods.2.Customer Insights: Understanding customer behavior is vital for businesses to develop effective marketing strategies and enhance customer satisfaction. Data analytics can help businesses uncover insights into customer preferences, buying habits, and feedback, leading to personalized marketing campaigns and improved customer experiences.3.Operational Efficiency: Data analytics can streamline business operations by identifying inefficiencies and suggesting improvements. For instance, analyzing supply chain data can help organizations optimize inventory levels and reduce costs.4.Predictive Modeling: Predictive analytics, a subset of data analytics, involves using historical data to make predictions about future events. This can be particularly useful in industries such as finance, healthcare, and retail, where anticipating future trends can lead to competitive advantages.Questions1.What is the primary purpose of data analytics in business decision-making?A. To enhance creativity and innovation.B. To base decisions on factual evidence.C. To eliminate the need for research.D. To provide entertainment for employees.2.According to the passage, how can data analytics benefit customer satisfaction?A. By reducing customer interaction.B. By providing personalized marketing campaigns.C. By increasing the number of competitors.D. By decreasing customer feedback.3.Which of the following is NOT mentioned as an aspect where data analytics can improve business operations?A. Supply chain management.B. Marketing strategies.C. Employee training.D. Inventory optimization.4.What is the main advantage of predictive analytics over traditional analysis methods?A. It requires less historical data.B. It can be used for a wider range of industries.C. It provides more accurate predictions.D. It is less time-consuming.5.Why is data analytics becoming increasingly important in modern business?A. Due to the decline in data availability.B. Due to the rise in data volume.C. Due to the decrease in skilled professionals.D. Due to the elimination of traditional analysis methods.Answers1.B. To base decisions on factual evidence.2.B. By providing personalized marketing campaigns.3.C. Employee training.4.C. It provides more accurate predictions.5.B. Due to the rise in data volume.四、翻译（本大题有5小题，每小题3分，共15分）第一题Translate the following passage into English.原文：“随着互联网的普及，人们获取信息的渠道越来越多样化。

Refactoring OCL Annotated UML ClassDiagramsSlaviˇs a Markovi´c and Thomas Baar´Ecole Polytechnique F´e d´e rale de Lausanne(EPFL)School of Computer and Communication SciencesCH-1015Lausanne,Switzerland{slavisa.markovic,thomas.baar}@epfl.chAbstract.Refactoring of UML class diagrams is an emerging researchtopic and heavily inspired by refactoring of program code written inobject-oriented implementation languages.Current class diagram refac-toring techniques concentrate on the diagrammatic part but neglectOCL constraints that might become syntactically incorrect by chang-ing the underlying class diagram.This paper formalizes the most im-portant refactoring rules for class diagrams and classifies them with re-spect to their impact on attached OCL constraints.For refactoring rulesthat have an impact on OCL constraints,we formalize the necessarychanges of the attached constraints.Our refactoring rules are specifiedin a graph-grammar inspired formalism.They have been implementedas QVT transformation rules.Wefinally discuss for our refactoring rulesthe problem of syntax preservation and show,by using the KeY-system,how this can be resolved.Keywords:Refactoring,QVT,Imperative OCL,Graph-transformations,Syn-tax preserving refactoring rules,Source code verification1IntroductionModern software development processes,such as Rational Unified Process(RUP) [17]and eXtreme Programming(XP)[6],propagate the application of refactoring to support iterative software development.Refactoring(see[20]for an overview) is a structured technique to improve the quality of software artifacts.Artifacts produced in all phases of the software development life cycle could become a subject of refactoring.Existing techniques and tools,however,still mainly target implementation code.An up-to-date list of existing tools and their application domains can be found in[30].In his seminal work[28],Opdyke has introduced the concept of refactoring of implementation code.He defined refactorings as”...reorganization plans that support change at an intermediate level”and identified26of such reorganization This work was supported by Swiss National Scientific Research Fund under reference number2000-067917.plans;now better known as refactoring rules.A refactoring rule for implemen-tation code describes usually three main activities:1.Identify the parts of the program that should be refactored(code smells).2.Improve the quality of the identified part by applying refactoring rules,e.g.the rule MoveAttribute moves one attribute to another class.As the result of this activity,code smells such as LargeClass disappear.3.Change the program at all other locations which are affected by the refactor-ing done in step2.For example,if at some location in the code the moved attribute is accessed,this attribute call became syntactically incorrect in step2and must be rewritten.The application of refactoring rules,called refactoring steps,is most often pattern-driven.A design,that is an instance of Design Patterns[14]can usually be extended and maintained much better than a design that is less structured. Thus,it is often beneficiary to transform the current design towards an instance of Design Patterns.Pattern-driven refactoring steps have been thoroughly stud-ied in[16,21].To a certain degree,refactoring rules depend on the language they are applied on.This explains why there are many catalogs of refactoring rules for different languages.The most complete and influential catalog was published by Fowler in[13]for the refactoring of Java code.The refactoring of artifacts that are more abstract than implementation code is a relatively new research topic that became urgent with the success of the UML.Some initial catalogs of refactoring rules for UML diagrams are presented in[2,31,33].However,neither these catalogs nor any of the existing UML refactoring tools[8,9,29]support–apart from some simple Rename-refactorings–the refactoring of attached OCL constraints once the underlying UML class diagram has changed.Speaking in terms of the above given MoveAttribute example,thefirst two steps have been realized by many tools but the last step is usually ignored.For the refactoring of OCL,we are aware of only one approach.Correa and Werner present in[12]some refactoring rules for OCL,but these rules focus on the improvement of poorly structured OCL constraints and take only to a very limited extent the relationship between OCL constraints and the underlying class diagram into account.We give in this paper a formal specification of the most important refactor-ing rules for UML class diagrams including the necessary changes on attached OCL constraints.Trivially,it is always necessary to change an OCL constraint if the refactoring of the underlying class diagram would make this constraint syn-tactically invalid.We believe that our rules are syntax preserving,i.e.each rule preserves the syntactical correctness of the UML/OCL model it is applied on. Section2contains some’design guidelines’for the development of syntax preserv-ing refactoring rules.However,we have formally proved the syntax preservation property only for the rule ExtractClass(see Sect.5).Another criterion for refac-toring rules is semantics preservation,i.e.the meaning(semantics)of the model remains the same whenever the refactoring rule is applied.We do not discuss the problem of semantics preservation in this paper,but refer the interested reader to[4].Our formal description of refactoring rules is done on the level of the meta-model for UML and OCL.Unlike other approaches that describe refactoring rules formally[12,15,31,33],we do not use OCL pre-/postconditions for this purpose.The formalism of our choice is the graph-grammar inspired notation proposed by the QVT Merge Group in an early submission for the OMG standard Query/View/Transformation(QVT)[25].Unfortunately,the graphical notation for QVT has changed in thefinal adopted specification[26],but we keep using the initial graphical notation in order to keep the rules presented in this paper similar to those presented in the conference version of this paper[18].All refactoring rules presented in this paper have been implemented using Together Architect2006for Eclipse,a commercial CASE tool that supports development and execution of QVT transformations[9].The implementation of our rules can be downloaded from[19].The paper is organized as follows.In Sect.2we give preliminaries necessary to understand our rules,which are formally defined in Sect.3.An overview of implementation steps is given in Sect.4.In Sect.5we show on one example how syntax preservation of refactoring rules can be proven formally.For this task, the KeY-system has been successfully applied.Section6concludes the paper. 2Formalizing Refactoring Rules in QVTModel transformations are widely recognized as the heart and soul of model-driven development[32].Refactoring rules are a special form of model transfor-mations for which the source and the target model are expressed using the same language.In this paper,we describe refactoring rules in a graphical formalism that has been suggested by the QVT Merge Group in[25].The following subsections give a brief introduction to the most important elements of QVT.We discuss using a very simple example the general structure of refactoring rules and give some guidelines for achieving syntax preservation. Finally,since our aim is to refactor UML class diagrams together with attached OCL constraints,we recall the relevant parts of the official UML/OCL meta-model in order to facilitate the understanding of refactoring rules presented in Sect.3.2.1Basic ConceptsA model transformation is defined as a set of transformation rules.In the graphi-cal notation proposed in[25],a transformation rule consists of two patterns,LHS (left hand side)and RHS(right hand side),which are connected with the sym-bol.Optionally,a rule can have parameters and a when-clause containing a constraint written in OCL.The LHS and RHS patterns are denoted by a generalized form of object diagrams.In addition to the normal object diagrams,free variables can be used in order to indicate object identifiers and values of attributes.The same variable can occur both in LHS and RHS and refers at all occurrences–during the applicationof the rule–to the same value.In order to distinguish between objects/links occurring in patterns and objects/links occurring in concrete models we will use the terms pattern objects/links and concrete objects/links,respectively.A rule is applied on a source model(represented as an instance of the meta-model,i.e.as a graph)as follows:In the source model,a subgraph that matches with LHS is searched and rewritten by a new subgraph derived from RHS un-der the same matching.If the obtained target model still contains subgraphs matching with LHS,the rule is applied iteratively as long as it is applicable (possibly infinitely often).A matching is an assignment of all variables occur-ring in LHS/RHS to concrete values.When applying a rule,the matching must obey the restrictions imposed by the when-clause.This semantics of QVT rules has the following consequences:If a pattern object appears in the rule’s RHS but not in its LHS(i.e.,in LHS there is no pattern object identified by the same variable)then–when applying the rule–a corresponding,concrete object is created.If there is a pattern object in LHS but not in RHS,then the matching object in the source model is deleted together with all’dangling links’.Simi-larly,a link is created/deleted if the corresponding pattern link does not appear in both LHS and RHS(pattern links are identified by their role names and the pattern objects they connect).The value of an attribute for a concrete object is changed only if the attribute is shown on the corresponding pattern object in RHS.The attribute’s new value for the concrete object is obtained by the expression shown as value for the attribute in RHS under the current matching. Values of attributes that are not mentioned in RHS remain unchanged.2.2How to Write Syntax Preserving QVT RulesThe purpose of this subsection is twofold.Firstly,the section should illustrate on concrete examples the above given basic concepts of QVT,which already allow to write quite expressive transformation rules.Secondly,some basic principles of the design of syntax preserving refactoring rules are explained.These principles have been frequently applied for the design of the(more complex)rules presented in Sect.3for the refactoring of UML/OCL models.Example:Item-View World In order to explain QVT’s basic concepts,we start with refactoring rules for a tiny Item-View language,Fig.1shows its metamodel.There are two non-abstract language concepts Item and View,which both in-herit the metaattribute name from ModelElement.The metaassociation between Item and View indicates that arbitrarily many views can be attached to one item(which is called the owner of the view).Furthermore,the self-association on Item indicates that each item can have an arbitrary number of parent-and child-items.Moreover,we assume that the parent-child relationship is acyclic. This can be expressed in OCL by the following invariant:context Item inv C y cleFre e:s e l f.a l l P a r e n t s()−>e x c l u d e s(s e l f)Fig.1.Metamodel for simple Item-View languagecontext Item def:a l l P a r e n t s():Set(Item)=s e l f.parent−>union(s e l f.p a r e n t.a l l P a r e n t s()−>a s S e t()) Please note that the additional query allParents(),which represents the tran-sitive closure of the parent-relationship,is well-defined despite its recursive def-inition(see[3]for a detailed justification).Two Simple Refactoring Rules As afirst example,the renaming of an item, which has been selected by the user,is formalized by the QVT rule RenameItem1 as shown in the left part of Fig.2.The selected item is passed as thefirst parameter of the rule and the rule’s LHS checks whether the passed item really exists in the source model(what should,trivially,be always the case).The pattern RHS is identical to LHS except for attribute name,whose value is set to newName,the rule’s second parameter.(a)Item selected by user(b)Item identified by nameFig.2.Two versions of RenameItem refactoring ruleA second version of the Rename-refactoring is formalized by rule RenameItem2 shown in the right part of Fig.2.Here,the item that should be renamed is de-termined by a match of its name with thefirst parameter of the rule(oldName). Applied on a given source model,this rule would iteratively make the following two steps as long as possible:(1)search for an item with name oldName in the current model and(2)rename the found item to newName.Please note that the application of this rule might not terminate if there is an item with oldName in the source model and newName is the same as oldName.Also thefirst ruleRenameItem1suffers from the same problem.We will see later,how termination problems can be avoided by adding a when-clause to the QVT rule.Checking Syntax Preservation of a given Rule A refactoring rule is called syntax preserving if for every syntactically correct source model the obtained target model is syntactically correct as well.Syntactically correct models are exactly the valid instances of the metamodel,what boils down to the following three criteria:(1)all model elements are well-typed,(2)all multiplicity con-straints are met,and(3)all well-formedness rules are obeyed.Invalid instances of the Item-View metamodel would be,for example,an Item object having a value of type Integer for attribute name(fails to meet criterion(1)due to type declaration of name),a View object that is linked to two Item objects(see cri-terion(2)and multiplicity for owner),and an Item object having a self-link for association Inheritance(see criterion(3)and well-formedness rule CycleFree).If the syntax preservation of a given refactoring rule should be shown,it has to be argued for every valid metamodel instance that the refactoring rule is either not applicable or that the target model is a valid metamodel instance as well.Fortunately,only a single step of the rule application has to be taken into account.By a simple induction argument,one can lift the syntax preservation property from a single step to the whole rule application.The argumentation on the syntactical correctness of each possible target model can be split according to the three validity criteria given above.A detailed argumentation for the refac-toring RenameItem1is given in Table1.More generally,the following aspects should to be taken into account:–The target model is well-typed whenever RHS is well-typed.Note that ill-typed model elements can only stem from ill-typed pattern elements.The type correctness of RHS is,however,checked mechanically once the rule is implemented with a QVT editor such as Together Architect2006.–Multiplicity constraints should always be checked carefully whenever the rule creates or deletes objects/links.Please note that also all multiplicities from inherited associations have to be obeyed.–Arguing about the preservation of well-formedness rules requires the most effort.In afirst step,one has to identify all those well-formedness rules of the metamodel that might be affected by the refactoring.We have done this task for all UML/OCL refactorings manually,but,recently,an interesting approach to automate thisfiltering has been developed by Cabot[10,11].In a second step,convincing arguments have to be found that thefiltered well-formedness rules are obeyed in all possible target models.We show in Sect.5on one example,how such an argumentation can be formalized by using the KeY-system.Using when-clauses to Ensure Syntax Preservation The argumentation on the preservation of well-formedness rules is not always as trivial as the oneTable1.Arguing on the syntax preservation of RenameItem1In the RHS of the rule,all pattern objects,their Type Declarations attribute values and links between them are well-typed according to the metamodel.Since neither objects nor links are created/deleted Multiplicities by the rule application,all multiplicity constraintsare automatically obeyed in the target model.The only well-formedness rule is CycleFree andthe only change on a model that could make itinvalid is adding links for the Inheritance association Well-formedness Rulesto the model.Since this does not happenin the Rename-rules,the invariant CycleFree ispreserved.for RenameItem1shown in Table1.Often,a refactoring rule can(potentially) destroy many of the metamodel’s well-formedness rules.In this case,we need a more sophisticated argumentation why the refactoring rule is nevertheless syntax preserving.To illustrate the problem,we add another invariant to the Item-View metamodel:context Item inv UniqueNameInInheritance:s e l f.a l l P a r e n t s().name−>e x c l u d e s(s e l )Informally speaking,this well-formedness rule requires the name of each Item object to be different from the name of all its(transitive)parents.Obviously, this well-formedness rule is not always preserved by RenameItem1since there is no provision made to ensure that newName is not already used by any of the parents.This problem can befixed by using QVT’s when-clause.Afirst(not fully successful)attempt to correct the rule RenameItem1is shown in Fig.3.Fig.3.Renaming of selected item–when-clause is not sufficient to preserve Unique-NameInInheritanceThe when-clause adds some new restrictions for the application of the rule. The rule is only applicable on those subgraphs of the source model that(1) match with LHS and(2)for which the expression given in the when-clause is evaluated to true.Note that identifiers for pattern objects(here it)can be used within the rmally speaking,the rule is now only applicable on such Item objects whose parents have not already used newName as a name.Unfortunately,this when-clause does not preserve UniqueNameInInheritance in all cases.For example,suppose the rule is applied on concrete Item object it1whose parents have names different from newName.After the rule has been executed(and it1has been renamed to newName),the well-formedness rule is indeed valid for it1.However,it might be the case that the source model contains another object it2,which is a child of it1and which has the name newName as well.Then,UniqueNameInInheritance does not hold anymore in the target model for it2,because it has now the same name as its parent it1.In order to prevent such cases,the when-clause has to check not only for the parent items but also for the child items whether newName is already used as a name.The following additional operations facilitate to write the necessary when-clause in a compact way.context Item def:a l l C h i l d r e n():Set(Item)=s e l f.c h i l d−>union(s e l f.c h i l d.a l l C h i l d r e n()−>a s S e t())context Item def:a l l C o n f l i c t i n g N a m e s():Bag(S t r i n g)=s e l f.a l l P a r e n t s().name−>union(s e l f.a l l C h i l d r e n().name)−>i n c l u d i n g(s e l )The corrected version of the RenameItem1refactoring is shown in Fig.4. Note that the rule is applicable at most once and,thus,termination of the rule application is always ensured.Actually,many refactoring rules for UML/OCL have a very similar when-clause because the UML metamodel contains quite a few well-formedness rules imposing unique names for model elements.Fig.4.Renaming of selected item–correct version for UniqueNameInInheritance2.3Extends-Relationship between QVT RulesAnother important concept of QVT is the extends-relationship between rules. The need for extensions of QVT rules is motivated by the next well-formedness rule:context Item inv DerivedViewName:s e l f.view−>f o r A l l(v|=’viewOf’.c o n c a t(s e l ))Informally speaking,DerivedViewName stipulates that all views attached to the same item must share the same name,which can be derived from the item’s name.Again,each of the above given RenameItem-refactorings would fail to pre-serve this invariant.Interestingly,there are now at least three possibilities to fix this problem.One possibility is to disallow renaming of Item objects if they have already a view attached.This is easily realized by extending the existing when-clause shown in Fig.4by and it.view->isEmpty().A second possibility is to delete all attached View objects when an Item object is renamed.The third possibility is to rename all attached View objects accordingly.Fig.5.Extension of RenameItem1Figure5shows the realization of the third possibility in form of an extension of RenameItem1.The new rule is called UpdateViewNames and is applied in the following way:Whenever a match for LHS of the extended rule(RenameItem1) is found,all its extensions(here UpdateViewNames)are applied on the cur-rent LHS-match as often as possible.Note that the patterns LHS/RHS from the extension rule can use elements from the extended rule.For example,the pattern object it:Item in LHS of RenameItem1refers for every match in the source model to the same model element as the pattern object it:Item in LHS of UpdateViewNames.The pattern object v:View in LHS of UpdateViewNames matches iteratively with any View object that is attached to it.The RHS of UpdateViewNames enforces to rewrite the name of all these View objects with the value’viewOf’.concat(newName).The when-clause in UpdateViewNames ensures the termination of the rule application.2.4Metamodel of UML/OCLWe present now all parts of the official metamodel for UML1.5and OCL2.0 that are relevant for the refactoring rules presented in Sect.3.Our refactoring rules are still based on UML1.5(and not on UML2.0,which was already the official UML version in time of writing this paper)for many reasons.First of all, we had the goal to stay as close as possible to the conference version of this paper [18],where initial versions of the refactoring rules from Sect.3are presented.More importantly,however,the implementation of our approach had topmost priority and we encountered numerous problems when trying to apply the QVT engine we used to repositories containing UML2.0models.The refactoring rules we formulate in Sect.3for UML1.5are also applicable in a very similar way to UML2.0models.In Sect.3.3,we describe a possible migration process for refactoring rules from UML1.5to UML2.0.As an ex-ample,the UML1.5refactoring rules MoveAttribute and MoveAssociationEnd are merged to MoveProperty for UML2.0.So far,however,we were unable to implement this refactoring rule due to the above mentioned technical problems.Fig.6.UML-Core Backbone and RelationshipsDeclaration of Metaclasses Fig.6and Fig.7show relevant parts of the official metamodel for UML1.5and OCL2.0(for a complete definition see[22, 24]).The chosen fragment of the UML-part of the metamodel concentrates on the main concepts of class diagrams.The OCL-part covers the most important OCL expressions.Fig.7.OCL-Overview and PropertyCallExpWell-formedness Rules The metamodel for UML and OCL contains hundreds of well-formedness rules and,as we have seen above,each well-formedness rule can become crucial if the syntax preservation of the refactoring rule is discussed. Our refactoring rules are designed to preserve only some,but–as we believe –the most important well-formedness rules of UML/OCL.This decision was a trade-offbetween the completeness of our approach and the readability of when-clauses,which grow when more well-formedness rules have to be preserved.Since the UML/OCL refactorings considered in this paper mainly rename, move,or add model elements,the well-formedness rule ensuring the uniqueness of used names in a classifier is easily broken when the refactoring rules do not make any provision.According to the UML1.5metamodel,all attributes,opposite association ends and other owned elements(e.g.contained classes)of a classifier must have a unique name.Moreover,these names must also not be used by any of the parent classifiers.A(slightly simplified)version of the official well-formedness rule looks as follows:context C l a s s i f i e r inv UniqueUsedName:s e l f.allUsedNames()−>f o r A l l(n|s e l f.allUsedNames()−>count(n)=1)context C l a s s i f i e r def:allUsedNames():Bag(S t r i n g)=s e l f.a l l P a r e n t s()−>i n c l u d i n g(s e l f)−>i t e r a t e(c;acc:Bag(S t r i n g)=Bag{}|acc−>union(c.o p p o s i t e A s s o c i a t i o n E n d s().name)−>union(c.a t t r i b u t e s().name)−>union())As we have seen in the Item-View example,it is convenient to define an additional operation that will capture also the names already used in the children of a classifier.context C l a s s i f i e r def:a l l C o n f l i c t i n g N a m e s():Bag(S t r i n g)=s e l f.allUsedNames()−>union(s e l f.a l l C h i l d r e n()−>i n c l u d i n g(s e l f)−>i t e r a t e(c;acc:Bag(S t r i n g)=Bag{}|acc−>union(c.o p p o s i t e A s s o c i a t i o n E n d s().name)−>union(c.a t t r i b u t e s().name)−>union()))Please note that the definition of many additional operations such as Classifier.allParents():Set(Classifier),Classifier.allChildren():Set(Classifier),Classifier.conformsTo(Classifier):Boolean,etc.is omitted here but can be found in the official definition of the metamodel[22,24].A second important well-formedness rule in the metamodel of UML1.5is that two operations with the same signature can be owned by any two classifiers (even if one of the classifiers is a specialization of the other one),but the two operations cannot be owned by the same classifier.context C l a s s i f i e r inv UniqueMatchingSignature:s e l f.o p e r a t i o n s()−>f o r A l l(f,g|f.m a t c h e s S i g n a t u r e(g)i m p l i e s f=g)3A Catalog of UML/OCL Refactoring RulesIn this section,we present the most important refactoring rules for UML1.5 class diagrams.These rules handle OCL2.0constraints that are attached to the refactored class diagram.At the end of this section,in Subsection3.3,an example for a possible migration from refactoring rules for UML1.5to such for UML2.0 is given.Note that each refactoring rule is syntax preserving only with respect to the part of the UML1.5metamodel given in the last section.Some of the refactoring rules are designed also for the preservation of some further important well-formedness rules(encoding restrictions for OCL expressions)that are given in the text at appropriate places.Our catalog(see Table2for an overview)is inspired by the refactoring rules for the static structure of Java programs given by Fowler in[13].We took the free-dom to change some of the rule names introduced by Fowler in order to indicate UML as their new application domain(e.g.,MoveMethod became MoveOpera-tion).In few cases,not only the name but also the semantics of the rule haschanged(e.g.,PullUpOperation moves in our version only the selected operation whereas in[13]also relevantfields are moved).Not all class diagram refactoring rules have an influence on attached OCL constraints.Table2classifies the rules according to this criterion.Note that Rename-refactorings require to change the textual representation of relevant constraints but not their metamodel-representation.Table2.Overview of UML/OCL refactoring rulesRefactoring rules Influence on syntacticalcorrectness of OCL constraintsMM-Representation Textual Notation RenameClass No YesRenameAttribute No YesRenameOperation No YesRenameAssociationEnd No YesPullUpAttribute No NoPullUpOperation No NoPullUpAssociationEnd No NoPushDownAttribute*No NoPushDownOperation*No NoPushDownAssociationEnd*No NoExtractClass No NoExtractSuperclass No NoMoveAttribute Yes YesMoveOperation Yes YesMoveAssociationEnd Yes Yes*only push down to one subclass is considered in this paper3.1Rules Without Influence on OCLRenameClass/Attribute/Operation/AssociationEnd These rules are very similar to each other and only RenameAttribute(see Fig.8)is discussed here in detail.The Rename-rules differ mostly in the when-clause,whose purpose is to check whether the proposed new name is already in use in the enclosing Namespace of the renamed element.In rule RenameAttribute,the parameter a refers to the attribute whose name should be changed.Since RenameAttribute is designed to work on class diagrams, we make in LHS the assumption that the owner of a is a Class,though it could be any Classifier according to the metamodel(similar assumptions are made also in all other refactoring rules).Thefirst line of the when-clause is necessary to guarantee termination when applying the rule.The second line ensures the applicability of the rule only in cases,in which the new name of the attribute is not already used within the owning class or one of its parents or children.。

In the pursuit of excellence, the concept of hard work is paramount. It is the driving force that propels individuals to achieve their goals and overcome obstacles. Here is an essay on the importance of hard work, particularly in the context of learning English. The Significance of Hard Work in Learning EnglishEnglish has become a global language, essential for communication, business, and education. As such, mastering it is no small feat, but rather a journey that requires dedication, patience, and above all, hard work. The process of learning English is akin to climbing a mountain it is arduous, yet the view from the summit is worth every step.Firstly, hard work is the foundation of language acquisition. It is through consistent practice that one hones their listening, speaking, reading, and writing skills. For instance, immersing oneself in English mediasuch as films, music, and literatureprovides a rich environment for vocabulary expansion and idiomatic expression. This immersion is not a passive activity but requires active engagement and the willingness to learn from every encounter with the language.Secondly, hard work in learning English is characterized by setting and achieving milestones. Whether its passing an English proficiency test, delivering a presentation in English, or simply conversing fluently, each goal requires a structured plan and unwavering commitment. Setting realistic targets and breaking them down into manageable tasks can make the learning process more organized and less overwhelming. Moreover, the role of hard work extends to overcoming setbacks. Language learning is fraught with challenges, such as understanding complex grammar rules or grappling with pronunciation difficulties. It is through persistent effort and the courage to make mistakes that learners refine their skills. Embracing failure as a stepping stone to success is a testament to the power of hard work.In addition, hard work in the context of English learning often involves seeking out resources and opportunities for improvement. This could mean enrolling in language courses, joining conversation clubs, or utilizing language learning apps. The initiative to explore various avenues of learning is a clear demonstration of ones commitment to hard work.Lastly, the benefits of hard work in learning English are manifold. It not only enhances ones ability to communicate effectively but also broadens cultural horizons, opens up jobopportunities, and fosters cognitive development. The sense of accomplishment that comes from mastering a new language is a reward in itself, reinforcing the value of hard work.In conclusion, hard work is the cornerstone of successfully learning English. It demands time, effort, and resilience but offers a world of opportunities in return. As with any worthwhile endeavor, the journey is as important as the destination, and the process of learning English is a testament to the power of perseverance and dedication.。

2025年研究生考试考研英语（一201）自测试卷与参考答案一、完型填空（10分）Section I: Cloze Test (20 points)Directions: Read the following text. Choose the best word(s) for each numbered blank and mark A, B, C, or D on your answer sheet.Passage:In the world of higher education, there is an ongoing debate about the significance of standardized tests, particularly the Graduate Record Examinations (GRE) for those aspiring to pursue graduate studies. The GRE, particularly its English Language Test, known as GRE General Test (Verbal Reasoning), aims to assess a candidate’s ability to comprehend, analyze, and evaluate complex written materials. This section, specifically, the “Verbal Reasoning - Text Completion” segment, exemplifies this objective by presenting a passage with 20 blanks, each requiring a precise word or phrase to maintain the flow and meaning of the text.Text:Academic research is a meticulous process that demands not only a deep understanding of a subject matter but also the (1)_____to questionestablished knowledge and seek new perspectives. Researchers are often(2)_____with vast amounts of data, requiring them to possess excellent(3)_____skills to sift through and organize information effectively. The ability to (4)_____conclusions from such data is crucial, as it forms the basis of scientific discoveries and scholarly contributions.However, the path to research excellence is rarely (5) _____. It is fraught with challenges, including the pressure to publish in high-impact journals, the (6)_____for funding, and the constant need to innovate and stay (7)_____with the latest research trends. Despite these obstacles, researchers persevere, driven by their (8)_____to uncover the truth and make a meaningful impact on their fields.Collaboration is a cornerstone of the research process. Working together, researchers can pool their expertise, share resources, and (9)_____each other’s strengths and weaknesses. This not only accelerates the pace of research but also fosters an environment of (10)_____and mutual respect. In the realm of language and literature, researchers engage in critical analyses of texts, examining their (11)_____meaning, cultural context, and historical significance. The GRE English Test assesses this ability by testing cand idates’ comprehension of complex texts and their capacity to draw (12)_____from them. For instance, candidates may be asked to identify the author’s (13)_____or the tone of a passage, or to infer the implications of a statement made within the text.To excel in this section, candidates must develop a (14)_____vocabulary,enabling them to comprehend a wide range of vocabulary and idiomatic expressions. Additionally, honing their reading comprehension skills, such as identifying the main idea, supporting deta ils, and author’s purpose, is essential. Furthermore, the ability to (15)_____logical connections between ideas and sentences within a text is key to accurate interpretation. While preparing for the GRE, it is important to engage in regular practice, utilizing a variety of resources that mimic the actual test format. This includes working through (16)_____passages, analyzing their structure, and practicing answering questions similar to those found on the test. By doing so, candidates can (17)_____their skills and gain confidence in their abilities.Ultimately, the GRE English Test is a tool that measures a candidate’s readiness for graduate-level studies in the English language and literature. It is not a definitive measure of one’s intellectual capacity bu t rather an indication of one’s ability to navigate and excel in academic environments that prioritize (18)_____and critical thinking. As such, it is important for candidates to approach the test with a mindset focused on demonstrating their strengths and areas of improvement, rather than (19)_____on a single score.In conclusion, the GRE English Test is a challenging yet essential component of the graduate admissions process. By (20)_____a comprehensive preparation strategy that includes regular practice, vocabulary enhancement, and thedevelopment of critical reading skills, candidates can position themselves to succeed on this important milestone in their academic journey. Answers:1.courage2.confronted3.analytical4.draw5.smoothpetition7.current8.passion9.recognize10.collaboration11.literal12.inferences13.perspective14.robust15.establish16.practice17.refine18.research19.fixating20.adopting二、传统阅读理解（本部分有4大题，每大题10分，共40分）Section II: Traditional Reading ComprehensionFirst PassageTitle: The Digital Revolution and Its Impact on EducationIn the past few decades, the world has witnessed an unprecedented digital revolution that has transformed virtually every aspect of our lives. From the way we communicate to the way we access information, technology has played a pivotal role in shaping our societies. This transformation is particularly evident in the field of education, where the integration of digital tools and resources has not only revolutionized teachingmethodologies but also expanded learning opportunities for studentsglobally.The advent of the internet and the subsequent rise of online platforms have made educational resources accessible to millions of people who might not have had access to traditional educational institutions. Online courses, known as massive open online courses (MOOCs), have emerged as a popular mode of learning, offering a wide range of subjects from top universities around the world. These courses are often free or at a minimal cost, making them affordable for students from diverse economic backgrounds.Moreover, the use of digital tools in classrooms has enhanced the learningexperience for students. Interactive whiteboards, tablets, and educational software have made lessons more engaging and dynamic. Teachers can now incorporate multimedia elements such as videos, animations, and simulations into their lessons, making complex concepts easier to understand. Additionally, personalized learning programs, which utilize data analytics to tailor educational co ntent to individual students’ needs and strengths, are becoming increasingly common.However, the digital revolution in education has not been without its challenges. Concerns over digital addiction, privacy issues, and the potential for technological distractions in the classroom have been raised. Furthermore, the digital divide—the unequal distribution of access to technology and the internet—remains a significant barrier to achieving equitable education opportunities for all.Despite these challenges, the benefits of the digital revolution in education are undeniable. It has democratized access to knowledge, improved teaching and learning outcomes, and fostered innovation in educational practices. As technology continues to evolve, it is likely that the future of education will be even more deeply intertwined with digital tools and resources.Questions:1.What is the main topic of the passage?A)The rise of online shoppingB)The impact of the digital revolution on educationC)The history of the internetD)The challenges faced by traditional educational institutions2.Which of the following is NOT mentioned as a benefit of online courses?A)They are accessible to people from diverse economic backgrounds.B)They are taught by the best teachers in the world.C)They offer a wide range of subjects.D)They are often free or at a minimal cost.3.How do digital tools enhance the learning experience for students in classrooms?A)By making lessons less engaging and dynamic.B)By incorporating multimedia elements into lessons.C)By eliminating the need for teachers.D)By making it difficult to understand complex concepts.4.What is a major concern related to the digital revolution in education?A)The high cost of educational software.B)The lack of access to technology and the internet for some students.C)The excessive use of paper in classrooms.D)The decline in the quality of traditional educational institutions.5.What does the author suggest about the future of education in relation to digitaltools and resources?A)They will become less important over time.B)They will continue to play a minor role in educational practices.C)They will be completely replaced by traditional methods.D)They will become even more deeply intertwined with education.Second Section: Traditional Reading ComprehensionPassage:Title: The Impact of Digitalization on the Traditional Book IndustryIn recent years, the digital age has transformed nearly every aspect of our lives, and the book industry is no exception. With the rise of e-books, audiobooks, and digital reading platforms, the traditional paper-based model of book publishing and distribution is facing unprecedented challenges. This transformation has sparked debates among readers, authors, publishers, and librarians alike, as they grapple with the implications of this digital shift.At the heart of the matter lies the convenience offered by digital formats. E-books, for instance, can be accessed instantly on a range of devices, from smartphones to tablets, eliminating the need for physical storage space and allowing for seamless cross-device reading. They are also often cheaper than their physical counterparts, appealing to readers on a budget. Additionally, the advent of cloud storage and online libraries has made it easier than ever to access and share vast collections of books.However, these benefits come at a cost. Many argue that the digitalization of books threatens the cultural significance and physicality of the printed word. Books have traditionally served as tactile objects, conveying a sense of ownership and permanence that cannot be replicated by a screen. Moreover, the disappearance of physical bookstores has had a profound impact on communities, reducing opportunities for social interaction andbrowsing-based discovery.Authors and publishers, too, have been affected. While digital platforms have opened up new avenues for reaching readers worldwide, they have also created a crowded and competitive marketplace where visibility can be difficult to achieve. Furthermore, concerns over piracy and the loss of control over how their work is presented and distributed have led some to question the value of embracing digital formats.Yet, despite these challenges, the book industry is adapting. Publishers are exploring innovative ways to integrate digital elements into physical books, such as augmented reality and interactive features, to enhance the reading experience. Meanwhile, libraries are embracing digital resources while maintaining their physical collections, recognizing the importance of both formats for diverse user needs.Ultimately, the future of the book industry lies in a delicate balance between the traditional and the digital. As readers continue to demand convenience and accessibility, it is essential that the industry evolves to meet these needs while preserving the cultural and physical value of books.Questions:1.What is the main topic of the passage?A)The advantages of e-books over traditional books.B)The impact of digitalization on the book industry.C)The role of libraries in the digital age.D)The future of book publishing.2.What is one benefit of e-books mentioned in the passage?A)They require more physical storage space.B)They are often more expensive than physical books.C)They can be accessed instantly on various devices.D)They cannot be shared easily with others.3.According to the passage, what has been a negative impact of the decline ofphysical bookstores?A)Increased competition among publishers.B)Decreased social interaction opportunities.C)Higher prices for e-books.D)Increased piracy of books.4.How have some publishers responded to the challenges of digitalization?A)By completely abandoning physical books.B)By embracing only digital formats.C)By integrating digital elements into physical books.D)By ignoring the changing market trends.5.What is the author’s overall stance on the future of the book industry?A)It will completely shift to digital formats.B)It will maintain its traditional form without change.C)It requires a balance between the traditional and digital.D)It is impossible to predict its future trajectory.Third Question: Traditional Reading ComprehensionPassage:Title: The Digital Divide: Bridging the Gap in Access to TechnologyIn today’s rapidly evolving digital age, technology has become an integral part of our daily lives, shaping how we communicate, learn, work, and entertain ourselves. However, amidst this technological boom, a significant disparity exists –the digital divide, a term coined to describe the unequal distribution of access to, use of, or impact of information and communication technologies (ICTs) among individuals, households, businesses, and geographic areas.The digital divide manifests itself in various forms, but a primary concern lies in the gap between those who have access to the latest technological advancements and those who are left behind. This disparity can be attributed to several factors, including economic status, education levels, age, gender, and geographical location. In developing countries, the digital divide is often exacerbated by infrastructural limitations and affordability issues, while in developed nations, it may be a result of digital illiteracy or a lack of motivation to adopt new technologies. The consequences of the digital divide are far-reaching and multifaceted. On an individual level, limited access to technology can hinder educational opportunities, limit career prospects, and isolate individuals from social networks. At a societal level, it can exacerbate economic inequalities, widen the achievement gap among students, and stifle innovation and progress.Efforts to bridge the digital divide have been ongoing for years, with governments, non-profit organizations, and private companies working together to provide access to technology for those in need. Initiatives such as e-learning programs, community technology centers, and low-cost devices aim to increase digital literacy and ensure that no one is left behind in the digital age.Despite these efforts, the digital divide remains a persistent challenge. As technology continues to evolve at an unprecedented pace, it is crucial that we remain vigilant in our efforts to ensure equitable access to its benefits. Only by bridging the gap in access to technology can we ensure that everyone has the opportunity to thrive in the digital age. Questions:1.What is the digital divide, and what does it refer to?Answer: The digital divide refers to the unequal distribution of access to, use of, or impact of information and communication technologies (ICTs) among individuals, households, businesses, and geographic areas.2.What are some of the primary factors contributing to the digital divide?Answer: Some of the primary factors contributing to the digital divide include economic status, education levels, age, gender, and geographical location.3.How can limited access to technology hinder educational opportunities?Answer: Limited access to technology can hinder educational opportunities by restricting access to digital resources, such as online courses and educational software, which can be vital for learning and development.4.What are some initiatives aimed at bridging the digital divide?Answer: Some initiatives aimed at bridging the digital divide include e-learning programs, community technology centers, and low-cost devices, which aim to increase digital literacy and ensure equitable access to technology.5.Why is it important to bridge the digital divide?Answer: Bridging the digital divide is important because it ensures that everyone has equal access to the benefits of technology, which can help to reduce economic inequalities, improve educational outcomes, and foster innovation and progress. Section IV: Traditional Reading ComprehensionPassage FourTitle: The Impact of Digital Technology on Reading HabitsIn the digital age, the way we consume information has undergone a profound transformation. From the traditional paper-and-ink books to the sleek electronic screens of tablets and smartphones, the advent of digital technology has reshaped our reading habits in ways that were once unimaginable. This passage delves into the various aspects of how digitalization has influenced our reading experiences, both positively and negatively.The convenience offered by digital devices cannot be overstated. With the tap of a finger, readers can access an endless library of books, articles, and news from anywhere in the world. Gone are the days of trudging to the bookstore or waiting for a book to arrive in the mail. The instant gratification of digital reading appeals to many, especially those with busylifestyles who value time efficiency. Moreover, the ability to customize reading settings such as font size, background color, and brightness levels caters to individual preferences, enhancing the overall reading experience. However, the shift towards digital reading has also raised concerns about its impact on comprehension and retention. Some studies suggest that reading from screens can lead to decreased attention spans and reduced ability to process information deeply. The constant distractions of notifications and social media alerts can further fragment our focus, making it challenging to fully immerse oneself in a book or article. Additionally, the lack of tactile feedback from physical pages and the absence of the traditional smell and feel of a book can diminish the emotional connection readers form with the content.Moreover, the proliferation of digital content has led to an explosion of information, much of which is of questionable quality. The ease of publishing online has democratized access to the written word but has also opened the floodgates to misinformation and clickbait. Navigating through this deluge of content can be overwhelming, and readers must develop critical thinking skills to discern fact from fiction.Despite these challenges, digital technology also presents new opportunities for reading and learning. Interactive e-books, for instance, incorporate multimedia elements like videos, animations, and quizzes that can enrich the learning experience and make complex concepts more accessible.Furthermore, personalized recommendation algorithms can curate tailored reading lists based on an individual’s interests and reading history, fostering a sense of discovery and exploration.In conclusion, the impact of digital technology on reading habits is multifaceted. While it has undeniably brought about convenience and new forms of engagement, it has also raised concerns about comprehension, attention, and the quality of information available. As we continue to navigate this digital landscape, it is essential to strike a balance between embracing the benefits of technology and preserving the essence of reading as a deeply personal and enriching experience.Questions:1.What is the main advantage of digital reading mentioned in the passage?•A) The ability to access an endless library of content instantly. •B) The tactile feedback from physical pages.•C) The lack of distractions from notifications and social media. •D) The improved comprehension and retention of information.2.Which of the following is a concern raised about digital reading?•A) The enhanced emotional connection readers form with the content. •B) The increased attention spans and ability to process information deeply.•C) The decreased attention spans and reduced ability to process information deeply.•D) The limited customization options for reading settings.3.What is the primary issue with the proliferation of digital content mentioned in thepassage?•A) The lack of accessible information for readers.•B) The overabundance of high-quality content.•C) The challenge of navigating through a deluge of information, including misinformation.•D) The ease of publishing traditional books.4.How do interactive e-books contribute to the reading and learning experience?•A) By reducing the emotional connection readers form with the content. •B) By limiting access to multimedia elements like videos and animations. •C) By enriching the learning experience and making complex concepts more accessible.•D) By decreasing the convenience of digital reading.5.What is the overall message of the passage regarding the impact of digitaltechnology on reading habits?•A) Digital technology has only negative impacts on reading habits. •B) Digital technology has completely replaced traditional reading methods.•C) The impact is multifaceted, with both positive and negative aspects that require balance.•D) The benefits of digital technology far outweigh any potential drawbacks.三、阅读理解新题型（10分）Title: The Rise of E-commerce and Its Impact on Traditional RetailIn recent years, the landscape of retail has undergone a dramatictransformation, fueled primarily by the exponential growth of e-commerce.Once a niche market, online shopping has now become an integral part of consu mers’ lives, challenging the dominance of brick-and-mortar stores.This shift has far-reaching implications, reshaping not only the way we shop but also the very fabric of our economic and social structures.The Convenience Factor: At the heart of e-commer ce’s success lies its unparalleled convenience. With just a few clicks, customers can browse through a vast selection of products from the comfort of their homes, compare prices effortlessly, and have their purchases delivered right to their doorsteps. This has not only saved time but also reduced the need for physical travel, making it especially appealing to busy professionals and those living in remote areas.Access to a Global Market: Another significant advantage of e-commerce is its ability to break down geographical barriers. No longer constrained by the limitations of their local markets, businesses can now reach customers worldwide. Similarly, consumers have access to an unprecedented range of products from across the globe, often at more competitive prices than those available locally.Challenges for Traditional Retail: However, this digital revolution has notcome without its challenges for traditional retailers. The rise ofe-commerce has led to a decline in footfall at physical stores, impacting sales and profitability. To stay afloat, many retailers have had to adapt by investing in their online presence, offering click-and-collect services, and enhancing in-store experiences to attract customers.The Future of Retail: The future of retail is likely to be a blend of both online and offline experiences, with retailers leveraging technology to create seamless omnichannel strategies. Augmented reality, virtual try-ons, and personalized recommendations are just a few examples of how technology is reshaping the shopping experience. As consumers continue to embrace digital solutions, retailers must innovate and evolve to meet their changing needs.Questions:1.What is the main driver behind the transformation of the retail landscape in recentyears?A)The increasing popularity of mobile payments.B)The exponential growth of e-commerce.C)The decline of physical infrastructure.D)The introduction of new tax policies.Answer: B2.Which of the following is NOT mentioned as an advantage of e-commerce?A)The ability to compare prices easily.B)The elimination of physical travel for shopping.C)Access to exclusive products not available locally.D)The convenience of shopping from home.Answer: C3.What does the term “omnichannel strategies” refer to in the context of retail?A) A single sales channel used by retailers.B) A blend of online and offline shopping experiences.C) A marketing technique focused on social media.D) A strategy to reduce operating costs.Answer: B4.How has the rise of e-commerce impacted traditional retailers?A)It has led to an increase in their sales and profitability.B)It has made them more competitive in the global market.C)It has caused a decline in footfall at their physical stores.D)It has made them completely obsolete in the retail industry.Answer: C5.Which technology is mentioned as having the potential to reshape the shoppingexperience?A)Artificial Intelligence.B)Augmented Reality.C)Internet of Things.D)Blockchain.Answer: B四、翻译（本大题有5小题，每小题2分，共10分）First QuestionQuestion: Translate the following paragraph into Chinese:The digital era has revolutionized the way we interact with information, making it possible to access vast amounts of knowledge instantly from anywhere in the world. This paradigm shift has not only altered our personal lives but also transformed industries, businesses, and the very fabric of society. As individuals navigate this ever-evolving landscape, it becomes increasingly crucial to develop a strong sense of digital literacy, enabling us to critically evaluate information, protect our privacy, and harness the power of technology for positive outcomes.Answer:数字时代彻底改变了我们与信息互动的方式，使我们能够瞬间从世界任何地方获取大量知识。

1cf.[Gel98]for an overview of the dynamic paradigm in cognitive science.1Proceedings of the Eighth ESSLLI Student SessionBalder ten Cate(editor)Chapter1,Copyright c2003,Alessandro MazzeiFormalizing a Constituency Based Dynamic GrammarIn the schema proposed by Milward([Mil94])to define a dynamic grammar,we have to define the states of the dynamic process,a set of axioms that relate a lexical item with a transition between states,some deduction rules that specify the time evolution of the process,and the sets of initial andfinal states.In this schema a grammar derives a sentence if and only if starting from an initial state and follow-ing the transitions associated with the words in the sentence,the decision process will be in afinal state after reaching the last word of the sentence.We can also specify a dynamic grammar starting from a lexicalized grammatical formalism, and redefining the combination operations in a dynamic fashion,as the axioms of the dynamic grammar.In such a way Milward defined the dynamic version of dependency grammars.In this paper,we formalize a TAG-related dynamic grammar,called Dynamic Version of TAG(DV-TAG),that satisfies strong incrementality.Following the in-formal work presented in[LS02],this dynamic grammar builds upon the lexicon and the attachment operations of Lexicalized TAG(L–TAG[Sch90])in order to define the states and the transitions.This paper is an ongoing work about viability of the formalism in term of basic mechanisms of syntactic construction.States are partially derived trees that span the input string from the beginning to the i-th word;transitions extend the left contexts by attaching some elementary tree anchored by the word through dynamic versions of Substitution and Adjoining. To our knowledge,DV-TAG is thefirst attempt to define a constituency based dy-namic grammar.The constituency nature of DV-TAG also allows to state easy relationship with the mainstream frameworks of computational linguistics.For ex-ample,we can introduce some statistical parameters and define a stochastic version of our formalism to yield a wide coverage parser.The following purely syntactic question remains open:where are DV-TAG grammars in the Chomsky hierarchy?We will show,as a starting point,that DV–TAG is powerful enough to derive a linguistically relevant context–sensitive lan-guage.Many other linguistic phenomena,including extractions and nested depen-dencies,can be accounted by DV-TAG.As an example we will show a derivation in DV-TAG for cross-serial dependencies.2Dynamic Version of Tree Adjoining Grammar(DV-TAG) The standard version L-TAG does not respect the strong incrementality:we cannot always build a fully connected partially derived tree composing the elementary trees and following the word sequential order.The intuitive idea behind DV-TAG is to convert the derivation process of Lex-icalized TAG into a dynamic system.A DV-TAG grammar,as L-TAG grammar ([Sch90]),is a set of elementary trees,divided into initial trees and auxiliary trees, and attachment operations for combining them.In DV-TAG attachment operations always occur between the left context and an elementary tree,and some combi-nations are not permitted.In particular,the derivation tree,which illustrates the2Alessandro Mazzeiderivation process in L-TAG(a context-free process),becomes a derivation se-quence.In the rest of this section,we sketch a formal definition of DV-TAG.This definition is necessary for the study of the expressive power of the formalism,that we will show in future works.We will not use linguistic examples,only formal symbols(a,b,c,etc.)without semantic role,since we want to stress the discussion about syntactic issues.Therefore the paper does not present a formal definition of semantic interpretation.First of all,we provide some general terminology.Let be an alphabet of ter-minal symbols,and an alphabet of non–terminal symbols.indi-cate terminal symbols,for non–terminal symbols, are strings of terminals,are mixed strings;is the lengthof the string.We denote initial trees with,auxiliary trees with,derived and generic trees with;with we denote a node belonging to a tree.is the string of terminal and non–terminal symbols on the frontier of a tree;is the–th element of the.We introduce two useful notions that are borrowed from parsing2dotted tree and fringe.Definition2.1.A dotted tree is a couple where is a tree and. Definition2.2.The left fringe of dotted tree is a path from to the root,minus the path from to the root.The right fringe of a dotted tree is the path from to root,minus the path from to root. We will use dotted lines to denote fringes in the trees(fig1.1).Dotted tree is the basic data structure in DV–TAG:the dot denotes a point in the yield of the tree and the fringes point where the next attachment operation can be applied.Now we can define the attachment operations.Since an elementary tree brings new lexical material in the string,we need to be careful about the linear positioning of such material during the derivation process,in order to avoid holes in the left fragment of the sentence.Constraints involve the shape of the elementary trees (in particular,we keep distinct three types of auxiliary trees)and the definition of operations.Following the definitions of[SW95],we define left auxiliary trees as auxiliary trees that have(non–empty)terminal symbols only on the left of the foot node,right auxiliary trees as auxiliary trees that have(non–empty) terminal symbols only on the right of the foot node,wrapping auxiliary treesas auxiliary trees that have(non–empty)terminal symbols on both the left and the right of the foot node.Elementary trees where the leftmost symbol of the yield is a terminal symbol are called left anchored trees.For left anchored auxiliary trees, it can be the case that the foot node is at the left of the leftmost anchor.We define six attachment operations on a dotted tree:two substitutions(similarFormalizing a Constituency Based Dynamic GrammarFigure1.1:Operations in DV–TAG.to L–TAG substitution),three adjunctions(similar to L–TAG adjunction),and one shift.SubstitutionLet be a left anchored initial tree that has the same root label as the substi-tution node N on the right fringe of.The root of is merged into the node N,and the dot in the new dotted tree is immediately to the right of the leftmost anchor of(figure1.1-A).Inverse SubstitutionLet be an initial tree,such that is a substitution node N labelled like the root of,and is the leftmost anchor of:the root of is merged into N,and the dot in the new dotted tree is on the right of the left anchor of(figure1.1-B).ShiftingLet be a dotted tree.The dot is shifted on the right to the next position4Alessandro MazzeiFigure1.2:Operations in DV–TAG.of the frontier(figure1.1-C).Adjoining from the leftLet be a left anchored left or wrapping auxiliary tree: is grafted into a non–terminal node that belongs to the right fringe of the dotted tree and has the same label as the root of.In the new dotted tree,the dot is at the position of the frontier(figure1.2-D). Adjoining from the rightLet be a left anchored right auxiliary tree:is grafted into a non–terminal node that belongs to the left fringe of the dotted tree and has the same label as the root of.In the new dotted tree the dot is in the position of the frontier(figure1.2-E).Inverse Adjoining from the Left5Formalizing a Constituency Based Dynamic GrammarLet be a left or wrapping auxiliary tree:is grafted into a non–terminal node that belongs to the right fringe of the dotted tree.In the new dotted tree the dot is to the right of the leftmost anchor of(figure1.2-F).Finally,we can give a formal definition of a DV–TAG.We adapt the schema defined in[Mil94]for the dynamic dependency grammars.Like in L–TAG,we have two sets of trees:a set of initial trees,and a set of auxiliary trees.Definition2.3.Given a lexicon,the DV–TAG G is a quadruple con-sisting of the set of states(including initial andfinal states),the set of axioms,the deduction rule,and the specification of the legal strings and trees.Set of states,where a state is a dotted tree;–Initial states:–Final states::.Set of axioms:1.2.,leftmost anchor of,with,or,orDeduction rule(a,b terminal symbols):Alessandro MazzeiFigure1.3:A fragment of DV–TAG derivation of the string“abcabc”.–A tree is derived by the grammar if and only if such thatwithfinal state.Left–context can be defined as a dotted tree such that, with,and.DV–TAG derivation is the sequence of left contexts.Cross–serial dependenciesFormally speaking,the dependencies in a string are equivalence relations between elements of the words.As usual we represent these relations with lines that link the elements.A grammatical system generates dependencies in a string,if in the derivation of the string,elements in the dependency relation are inserted in the same step of the derivation([BRN92]).The derivational generative capacity of a grammatical formalism is the set of the dependencies that the formalism can generate.Cross-serial dependencies,expressed by the string,are well-known test-bed for grammatical formalism attempting to describe natural languages.There7Formalizing a Constituency Based Dynamic Grammarare real linguistic constructions,as subordinate phrase construction in Dutch,that show these dependencies.It is easy to show that context-free grammars cannot generate this type of dependencies,therefore we need a more powerful formalism to take this phenomenon into account.Indeed the cross-serial dependencies play a key role in the definition of mildly context-sensitive languages([VSJW90]).We show that DV-TAG is powerful enough to derive cross-serial dependencies, and to produce a derived tree equal to the verb-raising analysis of Dutch sub-ordinate phrase construction given in[KS91],using a DV-TAG with the lexicon(figure1.3).The complete DV–TAG derivation of the string“abcabc”is:,where thefirst operation is an adjoining from the left operation,the second and third operations are inverse adjoining from the left operations(figure1.3depicts these operations),and the last six steps are shifting operations.Beyond the CF–powerWeakly generative capacity of DV-TAG is greater than weakly generative capacity of CF-grammars.This expressive power is a direct consequence of L–TAG gener-ative capacity([VSJW90]).DV–TAG with the lexicon(figure1.4)derives the languageThis language is the context-sensitive language3related to example discussed by Shieber([Shi85])against the context-freeness of natural languages.There is the initial fragment of the DV-TAG derivation for the sentence“”infigure1.4. 3Open questionsWe gave a formal definition for a dynamic grammar based on tree adjoining gram-mars,and we showed that several linguistic related formal languages can be derived by these grammars.In this paper we do not define any semantics for DV-TAG(we are stile working on this part of the formalism)even if there are two important issues that could be pointed out about semantics:In standard L-TAG we can couple each elementary tree with a semantic unit.Thus obtaining the semantics of the derived tree via the composition of theseAlessandro MazzeiFigure1.4:A DV–TAG deriving a context–sensitive language units.We have to distinguish adjoining and substitution through different rules of semantic composition.We can say that“the derivation trees repre-sent predicate-argument structures”([AR00]).In DV-TAG,in order to satisfy the strong incrementality,and then to allow incremental construction of a derivation tree,we have to“augment”the el-ementary structures with nodes that are not projected by the lexical anchor of the tree4.To respect the semantic compositionality we have to define an “unification-substitution”operation to check which nodes of the elementary tree are yet in the partially derived tree([LS02]).In this paper,we did not formalize this unification operation,because it does not change the expres-sive power of the formalism,even if it is necessary for linguistic reasons.In the standard TAG,the derivation tree plays a key role in the definition of semantic interpretation but,the way the derivation tree is built is irrelevant.In DV-TAG we have a more powerful structure,called derivation sequence, describing the derivation tree and also how it is built.In our opinion,seman-tics for DV-TAG could be better defined using this structure.In the future work we willfirst draw the exact position of DV–TAG in Chomsky hierarchy,answering the question whether DV–TAGs are mildly context-sensitiveBIBLIOGRAPHYgrammars([VSJW90]).In this case,we have to show that DV–TAG is polynomi-ally parsable and this will allow us to develop a wide coverage parser. Bibliography[AR00]Abeill´e Anne and Owen Rambow,editors.Tree Adjoining Grammars:For-malisms,Linguistic Analysis and Processing.(CSLI-LN)Center for the Studyof Language and Information-Lecture Notes,2000.[BRN92]T.Becker,O.Rambow,and M.Niv.The Derivational Power of Formal System or Scrambling is beyond LCFRS.Technical report,Institute for Research inCognitive Science,University of Pennsylvania,1992.[BW84]R.Berwick and A.Weinberg.The grammatical basis of linguistic perfor-mance:language use and acquisition.MIT Press,1984.[Gel98]Tim Van Gelder.The dynamical hypothesis in cognitive science.Behavioral and Brain Sciences,21:1–14,1998.[KS91] A.Kroch and B.Santorini.The derived constituent structure of the West Ger-manic verb-raising construction.In R.Freidin,editor,Principles and Param-eters in comparative grammar,pages269–338.MIT Press,Cambridge:MA,1991.[LS02]Vincenzo Lombardo and Patrick Sturt.Towards a dynamic version of tag.In Proceedings of6th International Workshop on Tree Adjoining Grammars andRelated Frameworks(TAG+6),pages101–110,Venice(Italy),20-23May2002.[Mil94] ward.Dynamic dependency grammar.Linguistics and Philosophy, 17(6),1994.[MW73]W.Marslen-Wilson.Linguistic structure and speech shadowing at very short latencies.Nature,244:522–533,1973.[Sch90]Yves Schabes.Mathematical and Computational Aspects of Lexicalized Grammars.PhD thesis,Department of Computer and Information Science,University of Pennsylvania,1990.[Shi85]S.M.Shieber.Evidence against the context-freeness of natural language.Lin-guistics and Philosophy,8:333–343,1985.[Sta94] E.P.Stabler.Thefinite connectivity of linguistic structure.In C.Clifton, L.Frazier,and K.Reyner,editors,Perspectives on Sentence Processing,pages303–wrence Erlbaum Associates,1994.[Ste00]M.J.Steedman.The syntactic process.A Bradford Book,The MIT Press, 2000.[SW95]Yves Schabes and Richard C.Waters.Tree insertion grammar:A cubic-time, parsable formalism that lexicalizes Context-free grammar without changingthe trees putational Linguistics,21(4):479–513,1995. [VSJW90]K.Vijay-Shanker,A.Joshi,and D.Weir.The convergence of mildly context-sensitive grammatical formalisms.In Peter Sells,Stuart Shieber,and Tom Wa-sow,editors,Foundational Issues in Natual Language Processing.MIT Press,Cambridge MA,1990.10。

PHYSICAL REVIEW A84,062306(2011)Multipartite entanglement detection from correlationJulio I.de Vicente1,*and Marcus Huber2,†1Institut f¨u r Theoretische Physik,Universit¨a t Innsbruck,Technikerstrasse25,A-6020Innsbruck,Austria 2Faculty of Physics,University of Vienna,Boltzmanngasse5,A-1090Vienna,Austria(Received6July2011;published6December2011)We introduce a general framework for detecting genuine multipartite entanglement and non-full-separability in multipartite quantum systems of arbitrary dimensions based on correlation tensors.Regarding genuine multipartite entanglement,our conditions are comparable to those of previous approaches in the case of qubits,while they show particular strength in the relatively unexplored case of higher-dimensional systems.In the case of non-full-separability,our conditions prove to be advantageous in situations where more than two-body correlations are relevant,where most previous conditions turned out to be weak.Moreover,they allow for the detection of fully bound entangled states.Finally,we also discuss experimentally friendly ways of implementing our conditions, which are based on directly measurable quantities.DOI:10.1103/PhysRevA.84.062306PACS number(s):03.67.Mn,03.65.UdI.INTRODUCTIONIn many-body quantum physics entanglement constitutes a fundamental plex systems with multipartite quantum correlations can be exploited to enable numerous tasks in quantum-information processing.The multipartite entanglement in these systems enables quantum computation (e.g.,[1]),multiparty cryptography(e.g.,[2–4]),and the implementation of various other quantum algorithms(e.g., [5]).Apart from these possible applications in modern quan-tum technologies,it has become apparent that multipartite entanglement also plays a fundamental role in the physics of complex systems.While the involvement in quantum phase transitions(e.g.,[6])and ionization procedures(e.g.,[7]) seems clear,the recently suggested role in biological systems is still a subject of debate(e.g.,[8–11]).Therefore,to decide if a state is entangled or not is a fundamental problem in quantum-information theory[12,13]. Although a simple mathematical characterization is elusive (the problem has in fact been proved to be nondeterministic-polynomial-time hard or NP hard[14]),several works have put up sufficient conditions to identify a multipartite state as entangled[15–18].These conditions not only are helpful for entanglement detection but also provide more physical insight into this phenomenon.In contrast to the bipartite case,there exist different classes of multipartite entangled states.Genuine multipartite entanglement is of particular interest since it involves entanglement between all the subsystems.Recently there has been a lot of progress concerning its detection, mostly using linear and nonlinear entanglement witnesses [19–27]and Bell-like inequalities[28,29].However,with a few exceptions(see,e.g.,[20]),the approaches taken in each particular case only allow detection of either entanglement or genuine multipartite entanglement.Moreover,most of them are limited to qubit systems.In this paper we develop a general framework which allows use of the same piece of information to detect both entanglement and genuinely multipartite entanglement for *julio.de-vicente@uibk.ac.at†marcus.huber@univie.ac.at multipartite states of arbitrary dimensions.Our main tool will be correlation tensors which are built from the expectation values of a local operator basis.Our motivation stems from different facts.First,it has been shown that all information about the entanglement properties of a system is encoded in the correlation tensors[30],and these mathematical tools have already been proven useful for the detection of entanglement in the bipartite case[31].In[16]afirst step toward the extension of these ideas to the detection of multipartite entanglement was taken(see also[32]for a correlation-tensor approach to multipartite entanglement detection).However,here we will show that this allows for a much more general formalism (in which the criterion of[16]is a particular case and that of[32]is strictly weaker),which,furthermore,enables us to identify different classes of multipartite entanglement. Our conditions are expressed through simple mathematical inequalities.In contrast to entanglement witnesses,which are designed for a particular class of states,violations of these inequalities signal genuine multipartite entanglement or non-full-separability for general states.Moreover,since the entries of the correlation tensors are directly related to measurable quantities,we will discuss how our approach can be adapted to optimize the experimental st,many conditions for multipartite entanglement such as spin squeezing inequalities [17],covariance matrices[18],entanglement witnesses based on structure factors[21]or two-particle Hamiltonians[33], or magnetic susceptibility measurements[34]rely only on two-body correlations.It has been shown in[18]that this limits their ability to detect entanglement as there are important classes of states like graph states which have the same two-particle reduced states as separable states.Hence,their entanglement cannot be revealed by just looking at two-point correlations.On the other hand,correlation tensors take into account all m-body correlations.This suggests(and we will later see)that correlation tensors may overcome the limitations of the previous criteria.II.PRELIMINARIESBefore we proceed to derive our main results let us briefly review the definitions of multipartite entanglement and correlation tensors.We consider n-partite quantum statesρJULIO I.DE VICENTE AND MARCUS HUBER PHYSICAL REVIEW A 84,062306(2011)acting on the Hilbert space H =H 1⊗···⊗H n of dimension D =d 1,d 2···d n .If a pure state | ∈H can be written as a tensor product of states for every subsystem,i.e.,| |=|ψ1 ψ1|⊗···⊗|ψn ψn |,(1)then the state is said to be fully separable.Consequently,fully separable mixed states are convex combinations of fully separable pure states.These states contain no entanglement at all .On the other hand,any n -partite pure state that can be written as a tensor product| |=| A A |⊗| ¯A ¯A |(2)with respect to some bipartition A ¯A(A denoting some subset of subsystems and ¯Aits complement)is called biseparable.These states might contain some entanglement (as | A and/or | ¯A might not be separable)but they are not completely entangled.States that are not biseparable with respect to any partition are then said to be genuinely multipartite entangled.The generalization to mixed states is straightforward.Any mixed state that can be decomposed into a convex sum of biseparable pure states is called biseparable.Consequently,any nonbiseparable mixed state is called genuinely multipartite entangled.Due to the fact that the bipartitions might differ for every element of the biseparable decomposition,it is an intricate task to find out whether such a decomposition is possible.Let {λ(j )i }d 2j−1i =1denote the generators of SU(d j )and let λ(j )0=I d j ,which altogether constitute an orthogonal basis of the real Hilbert-Schmidt space of Hermitian operators acting on H j [i.e.,with inner product A,B =Tr(AB )].Thus,so is { n j =1{λ(j )i }}for the operators acting on H and,hence,ρis completely characterized by the expectation values λ(1)i 1⊗···⊗λ(n )i n :=T i 1···i n where i j =0,1,...,d 2j −1,which gives rise to the so-called (multipartite)Bloch representation or (multipartite)Fano form of density operators.1We will decompose the tensor T i 1···i n into the m -body correlationtensors T (j )i j ,T (j,k )i j i k ,etc.,which are tensors of order m indicated by the number of labels in the superscript.All the indices not labeled in the superscript are fixed to be zero while the other indices take every possible value but zero (i.e.,the identity is not taken into account).For instance,the one-body correlation tensor for particle 1,given by T (1)i 1=T i 10···0with i 1=0,completely characterizes the reduced state ρ1,and thetwo-body correlation tensor for subsystems 1and 2,T (1,2)i 1i 2=T i 1i 20···0(i 1,i 2=0),together with the one-body correlation tensors of 1and 2,characterizes ρ12,and so on.For the n -body correlation tensor,which we shall also call the full correlation tensor,we will drop the superscripts to ease the notation,i.e.,T i 1···i n =T i 1···i n (i j =0∀j ).Given two tensors T i 1···i n and S j 1···j m ,their outer product ◦is the (n +m )th-order tensor (T ◦S )i 1···i n j 1···j m =T i 1···i n S j 1···j m .If some tensor can be written as the outer product of two other tensors,say T i 1···i n =R i 1i 2i 3S i 4···i n ,we will say that the tensor1Notice that T 0···0is fixed by the normalization condition Tr ρ=1and there are indeed j d 2j−1parameters.factorizes in the corresponding splitting ({1,2,3},{4,5,...,n }in this case).If a tensor cannot be written as the outer product of any two lower-order tensors,we will say that the tensor does not factorize.It has been shown in [31]that a bipartite pure state isseparable if and only if (iff)T (1,2)i 1i 2=T (1)i 1T (2)i 2.Accordingly,a multipartite pure state is biseparable with respect to thepartition A ¯A iff T (A,¯A )i A i ¯A =T (A )i A T (¯A )i ¯A .Thus,we have the following characterization of biseparable pure states:Fact 1.A pure state is biseparable iff there exists somepartition of the subsystems A ¯Afor which all the m -body correlation tensors involving k particles from A and m −kfrom ¯A(k =0,m )factorize into the corresponding k -body correlation tensor of the k particles from A and the (m −k )-body correlation tensor of the m −k particles from ¯A.This leads to a simple sufficient condition for genuinely multipartite entangled pure states:Corollary 1.If some m -body correlation tensor of a pure state cannot be factorized into meaningful lower-order correlation tensors,then the state contains genuine multipartite entanglement.Analogously,this can be extended to non-fully-separable states (see also [16]):Corollary 2.If some m -body correlation tensor of a pure state cannot be fully factorized into meaningful one-body correlation tensors,then the state is not fully separable.We stress that the factorization must be possible into meaningful correlation tensors.This a consequence of the fact that the Bloch representation holds for Hermitian operators and not only for density operators,which are furthermore positive semidefinite.Hence,not all values of T i 1···i n give rise to a density matrix,i.e.,are meaningful.To characterize this subset is a quite involved problem (see,e.g.,[35]).However,there exist several conditions that should be fulfilled by the set of meaningful correlations.For instance,it will be useful later on that for one-body correlation tensors it must hold that||T (j )||2(d j −1)d j ,(3)with equality iff the state is pure and where ||·||is the standard Euclidean norm for vectors.This expresses the factthat Tr ρ2j1.This condition holds for the following choice of normalization for the generators of SU(d j ):Tr(λm λn )=2δmn [of course m,n =0since for the identity we haveTr(λ(j )0λ(j )0)=d j ].We will follow this convention throughout the paper,with which for qubits the generators correspond to the standard Pauli matrices.Following [16,31],the main idea behind this paper is to express the factorizability of some tensor into lower-order meaningful tensors as an upper bound on some convex function.Convexity will then imply that this bound must hold as well for biseparable (fully separable)mixed states and,hence,a violation of this bound will signal the presence of genuine multipartite entanglement (non-full-separability)for general quantum states.It seems that some tensor norm is the best choice of convex function since the norm of meaningful correlation tensors is upper bounded,as we have just seen for one-body correlation tensors.Notice that convexity in this case is guaranteed by the triangle inequality.Physical intuitionMULTIPARTITE ENTANGLEMENT DETECTION FROM ...PHYSICAL REVIEW A 84,062306(2011)suggests that full correlation tensors should be the first ones to check and usually we will restrict ourselves to them.The rest of the paper is organized as follows.In Secs.III and IV we provide two different approaches that lead to conditions for the identification of genuine multipartite entanglement.In Sec.V we show that similar techniques can be used to obtain conditions for non-full-separability.Section VI is devoted to some mathematical properties of our conditions which are related to their experimental implementation.Final conclusions are drawn in Sec.VII .III.GENUINE MULTIPARTITE ENTANGLEMENTCONDITIONS BASED ON THE STANDARDTENSOR NORMThe standard tensor norm is defined as the natural gener-alization of the Euclidean vector norm to higher-order tensors (recall that we will always deal with real tensors),i.e.,||T i 1···i n ||2= i 1,...,i nT 2i 1···i n.(4)This seems to be a very good choice for our purposes since thisnorm is multiplicative under outer products,i.e.,||T ◦S ||=||T ||||S ||∀T ,S .Hence,we will just need to upper-bound the standard norm of the m -body correlation tensors.This turns out to be quite easy.As we mentioned above,the conditionthat Tr ρ2j 1must hold translates into an upper bound for the standard norm of the one-body correlation tensors.Now,combining this condition with Tr ρ2ij 1will yield an upper bound for the two-body correlation tensors (see,e.g.,[36]).This procedure can be recursively applied to upper-bound the standard norm of all meaningful m -body correlation tensors.For instance,this gives||T (j,k )|| 2d j d k −1d j d k ,(5)with equality iff the state ρjk is a maximally entangled state [36].To illustrate this,let us start by considering a tripartite pure state with subsystems of equal dimension d j =d ∀j .Then,full separability implies||T i 1i 2i 3||=||T (1)||||T (2)||||T (3)||=2(d −1)d 3/2,(6)and biseparability between any two subsystems and the other yields||T i 1i 2i 3||=||T (j )||||T (k,l )||8(d −1)(d 2−1)d 3.(7)Since the last condition is more restrictive and using convexity we then have the following.Theorem 1.If for an arbitrary (pure or mixed)tripartite state it holds that||T i 1i 2i 3||>8(d −1)(d 2−1)d 3,(8)then the state is genuinely multipartite entangled.This simple mathematical idea is already strong enough to detect paradigmatic cases of genuine multipartite entangle-ment.If we consider (8)for three qubits,this gives the bound √3 1.73while the Greenberger-Horne-Zeilinger (GHZ)and W states have respectively ||T i 1i 2i 3||=2and ||T i 1i 2i 3|| 1.92.Therefore,their genuine multipartite entanglement is successfully identified,leading to (modest)white noise tolerances of p GHZ 0.13and p W 0.10.2Furthermore,the power of this condition increases with the subsystem dimension improving remarkably on [23].In Fig.1we plot the detection power of Eq.(8)for dimensions 4,5,and 6.As the other criteria for genuine multipartite entanglement in high-dimensional systems are still based on qubit subsystems of the high-dimensional Hilbert space,it is perhaps not surprising that our criteria,exploiting all degrees of freedom,quickly outperform them with growing dimensionality of the system.As discussed above,the extension of this condition to states with more subsystems or different subsystem dimensions is straightforward.However,for four qubits we have for both the GHZ and the Dicke state with two excitations ||T i 1i 2i 3i 4||=3,which is precisely the same value of a tensor product of two maximally entangled bipartite states (||T i 1i 2i 3i 4||=√3×√3).On the analogy of (8),one might hope to improve for larger d ;nevertheless,in the next section we will present a different and more powerful approach.It is worth mentioning that ||T i 1···i n ||−[2(d −1)/d ]n/2has been shown to be an entanglement monotone [37].Our results show that a high value of this measure can imply not only some entanglement but even genuine multipartite entanglement.IV .GENUINE MULTIPARTITE ENTANGLEMENTCONDITIONS BASED ON NORMS OF MATRICIZATIONS OF TENSORSAs an alternative to the previous section one can seek for other norms.Unfortunately,to our knowledge,the standard norm is the only norm which is multiplicative under outer products,a property which is very convenient for the math-ematical simplicity of our derivations.Nevertheless,it turns out that considering matricizations of tensors (i.e.,particular rearrangements of the tensor values to form a matrix)[39]and the usage of matrix norms on these matricizations will lead to interesting and more powerful results.The matrix norms we will be dealing with are the Frobenius or Hilbert-Schmidt norm (which is the standard tensor norm on a matrix),the trace norm,and the Ky Fan k norms [40].That is,let A ∈R m ×n ;then||A ||=ijA 2ij =iσ2i ,||A ||tr =Tr √A T A =iσi ,(9)||A ||k =k i =1σi ,2Here and throughout the paper the white noise tolerance of someentanglement condition for a state ψis defined as the values of p for which p I D /D +(1−p )|ψ ψ|is still detected by this condition.JULIO I.DE VICENTE AND MARCUS HUBER PHYSICAL REVIEW A 84,062306(2011)FIG.1.(Color online)Here the parameter regions for which the state ρ=αρGHZ(d )+βρW (d )+1−α−β2d −2( d −1i =0|i,i,i +1 i,i,i +1|+|i +1,i +1,i i +1,i +1,i |)for (a)d =4,(b)d =5,and (c)d =6exhibits genuine multipartite entanglement are identified.The generalized GHZ and W states for d -dimensional systems are defined as ρGHZ(d )=|GHZ(d ) GHZ(d )|with |GHZ(d ) :=1√d d −1i =0|i,i,i and ρW (d )=|W (d ) W (d )|with |W (d ) =1√3(d −1)d −2i =0(|i,i,i +1 +|i,i +1,i +|i +1,i,i ).The (red)region labeled II uses criterion II from Ref.[23]optimized numerically over all local unitary representations of the density matrix.The (yellow)region labeled III uses criterion III from Ref.[23]optimized numerically over all local unitary representations of the density matrix.The numerical optimization was performed using the composite parametrization from Ref.[38].The (blue)region labeled C shows the states detected to be genuinely multipartite entangled using Eq.(8).where {σi }[i =1,...,min (m,n )]denote the singular values of the matrix,which are arranged,as usual,in nonincreasing order.Notice that the last Ky Fan norm is the trace norm,i.e.,||·||min(m,n )=||·||tr .We will define matricizations in the following way:(non)underlined indices are joined together in lexicographical order to give rise to the row (column)indices.Let A be thesubset of underlined indices;then we will call that an A,¯Amatricization.For example,let i j =1,...,n j ∀j ;thenT i 1i 2i 3i 4=⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝T 111k T 121k ···T 1n 21k T 112k T 122k ···T 1n 22k ...··...T 11n 3k ·······T 211k·······...··...T 21n 3k·······...··...T n 111k·······...··...T n 11n 3k ······T n 1n 2n 3k ⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠,(10)where T xyzk =(T xyz 1···T xyzn 4)is a row vector (k =1,2,...,n 4),is a 13,24matricization.In Dirac notation we would haveT i 1i 2i 3i 4=i 1,...,i 4T i 1i 2i 3i 4|i 1i 3 i 2i 4|.(11)This way of matricizing is a generalization of the concept of matrix unfolding or mode-n matricization that is often used in multilinear algebra [39],which corresponds to matricizations of one index giving rise to the row column and the rest to the column vectors.The matricizations we have defined are quiteconvenient for the problem at hand because they have a well-defined structure under the outer product of tensors.Of course,if all the indices of a tensor are joined together,the tensor is vectorized,T i 1i 2i 3=vec(T ),while it is straightforward to check that T i 1i 2W i 3i 4=(T i 1i 2)⊗(W i 3i 4).Concatenating these rules,a matrix form for any matricization of more involved outer products of tensors can be readily found.For instance,T i 1i 2i 3R i 4S i 5i 6W i 7i 8=(T i 2,i 1i 3)⊗(R i 4)T ⊗vec(S i 5i 6)⊗(W i 7i 8)T .(12)This is the kind of structure we need because the norms we are going to use are either multiplicative (||·||,||·||tr ,and ||·||1)or submultiplicative (||·||k )under tensor products.3So,analogously to the previous section,we just need to upper-bound these quantities for meaningful correlation tensors to obtain conditions for genuine multipartite entanglement.For the sake of simplicity we will consider multiqubit systems in the following sections.A.Three qubitsAccording to the above discussion the only thing left to be able to derive genuine multipartite entanglement conditions is to obtain upper bounds for the matrix norms of the correlation tensors,as we did with the standard norm in the previous section.We have the following.Lemma 1.The two-body correlation tensor of two qubits satisfiesT (j,l )i j i lkk ∀k (13)with equality iff the two qubits are in a maximally entangledstate.3This is a consequence of the fact that,if {σi }and {σ j }are respectively the singular values of the matrices A and B ,then the singular values of A ⊗B are {σi σ j }.MULTIPARTITE ENTANGLEMENT DETECTION FROM ...PHYSICAL REVIEW A 84,062306(2011)Proof.We will use the local unitary invariance of the norms of any matricization of the correlation tensors (see Sec.VI below).Notice then that,because SO(3) SU(2),the two-body correlation tensor (i.e.,correlation matrix)can be brought into diagonal form by choosing properly local unitaries in the two subsystems (see,e.g.,[41]).Since the entries of this matrix are expectation values of observables with eigenvalues 1or −1we have that |T ii | 1∀i .It can be readily checked that this bound is attained by the maximally entangled state (and only by the maximally entangled state because this value of the trace norm of the correlation matrix implies the maximal possible amount of entanglement,as measured,for instance,by the concurrence [42]). Lemma 2.If a pure three-qubit state is biseparable,then it holds that(i)if the state is fully separable||T i j i l i m ||k 1∀k ;(14)(ii)if the state contains no entanglement across j |lm||T i j i l i m ||k √∀k ;(15)(iii)if the state contains some entanglement across j |lm||T i j i l i m ||k k∀k.(16)Proof.We will use repeatedly the upper bounds (3),(5)and(13).(i)||T i j i l i m ||k = T (j )i j T (l )i lT (m )i mk =||(T (j ))·(T (l )⊗T (m ))T ||k=||T (j )||||T (l )⊗T (m )||=||T (j )||||T (l )||||T (m )||=1.(17)(ii)||T i j i l i m ||k = T (j )i j T (l,m )i l i mk =||(T (j ))·vec(T (l,m ))T ||k=||T (j )||||T (l,m )|| √3.(18)(iii)||T i j i l i m ||k = T (j,l )i j i l T (m )i mk =||(T (j,l ))⊗(T (m ))T ||k ||T(j,l )||k ||T(m )|| k.(19)From Lemma 2we read three sufficient conditions forgenuine multipartite entanglement,namely,that the norm of any of the three possible matricizations of the full correlation tensor is greater than √3,2,and 3for ||·||1,||·||2,and ||·||tr ,respectively.To illustrate the power of these conditions,consider that the singular values of these matricizations are {1.414,1.414,0}for the GHZ state and {1.374,0.943,0.943}for the W state.Hence,the last two conditions can detect genuine multipartite entanglement.Notice that these states are symmetric,so all matricizations of the full correlation tensor are equal;however,for general states our ability to detect a state as genuinely multipartite entangled might depend on the choice of matricization.To avoid this and to obtain a stronger condition which takes into account a combination of the bounds of Lemma 2rather than just picking one of them,we introduce the average matricization norm ||M (T i 1i 2i 3)||=(||T i 1i 2i 3||+||T i 1i 2i 3||+||T i 1i 2i 3||)/3,which leads to the follow-ing.Theorem 2.If for a three-qubit state it holds that||M (T i 1i 2i 3)||k >2k +√33,(20)then the state contains genuine multipartite entanglement.Proof.Simply use the fact that any biseparable state can be written as ρbs = k p k ρk 12⊗ρk 3+q k ρk 13⊗ρk 2+r k ρk 23⊗ρk1and combine properly the bounds (15)and (16).Thus,Theorem 2allows detection of genuine multipartite entanglement in mixtures of the GHZ and W states with white noise for noise levels of p GHZ 0.324and p W 0.209.Notice that it is known that these states are genuinely multipartite entangled iff p GHZ 0.571and p W 0.521[27].B.Four qubitsNow,as in previous sections,we need upper bounds to the norms of the matricizations of the three-body correlation tensor.However,it is not clear which states should attain the maximum values of these norms in opposition to the two-body case,where the maximally entangled state,as intuition would suggest,does the job.Moreover,numerics indicate that max |ψ ||T i j i l i m ||tr 3.272for a state which,although close to the W state,has no simple mathematical structure.This indicates that devising a systematic procedure to find the maximum value of these norms as we did with the standard norm in Sec.III might be very hard.Nevertheless,it turns out that we can use this procedure to obtain reasonable estimates by using the equivalence of the norms:||·||k √k ||·||.Lemma 3.The three-body correlation tensor of three qubits satisfies ||T (j,l,m )i j i l i m ||k 2√k ∀k .Proof.The result follows from ||T (j,l,m )i j i l i m || 2.To see thiswe proceed as in Sec.III .Tr ρ2jlm =1translates tos =j,l,m||T (s )||2+s<q||T (s,q )||2+||T (j,l,m )||2=7.(21)The minimum possible values of the norms of the lowest-order correlation tensors are ||T (s )||=0and ||T (s,q )||=1∀s,q (since the one-qubit reduced density matrices can be maximally mixed,but the highest mixing allowed by the two-qubit reduced density matrices is them being equal to the identity in a two-dimensional subspace).Therefore,||T (j,l,m )|| 2,which is attained by the GHZ state.Finally,notice that the Hilbert-Schmidt norm of any matricization of a tensor equals its standard norm as tensor.Notice that Lemma 3provides accurate estimates,as the trace norm bound 2√3 3.464is quite close to the numerical maximum given above,while the Ky Fan 2norm bound is actually sharp since it is attained by the GHZ state.Now,we can proceed as in Lemma 2to upper-bound ||T i j i l i m i s ||k .Then,for pure biseparable states,one obtainsthe bounds 2√k (for k 3and 2√3otherwise)if T i j i l i m i s =T (j )i j T (l,m,s )i l i m i s or T i j i l i m i s =T (j,l,m )i j i l i m T (s )i s(i.e.,the state is bisep-arable in one subsystem versus the other three,and we consider the two possibilities that the two indices giving riseJULIO I.DE VICENTE AND MARCUS HUBER PHYSICAL REVIEW A 84,062306(2011)to the row of the matricization belong to either unentangledor entangled particles),3if T i j i l i m i s =T (j,l )i j i l T (m,s )i m i sand k if T i j i l i m i s =T (j,m )i j i m T (l,s )i l i s.One could also consider one vs three matricizations of the full correlation tensor;however,one finds that ||T i j i l i m i s ||k √3k for T i j i l i m i s =T (j,l )i j i l T (m,s )i m i s (i.e.,any two vs two biseparable states),which turns out to be a weak condition and,thus,it is better not to take these matricizations into bining all the above bounds,defining the two vs two average matricization norm ||M 22(T i 1i 2i 3i 4)||=(||T i 1i 2i 3i 4||+||T i 1i 2i 3i 4||+||T i 1i 2i 3i 4||)/3,and proceeding as in Theorem 2,we have the following.Theorem 3.If for a four-qubit state one of the following inequalities holds:||M 22(T i 1i 2i 3i 4)||k >2√k,1 k 3,1+2k/3,4 k 9,(22)then the state contains genuine multipartite entanglement.With this,genuine multipartite entanglement is detected in the GHZ state with a white noise tolerance of p GHZ 0.307and for the Dicke states of one and two excitations we have respectively p D 1 0.018and p D 2 0.328.From [27]we know that there is genuine multipartite entanglement iff p GHZ 0.533and if p D 2 0.539.These examples indicate that the matricization approach is more powerful than that of the standard norm and that it can detect different classes of entangled states.Interestingly,the detection capability of Theorems 2and 3is already comparable to [23,26]for qubits,as shown in Fig.2.Using the matricization approach we have thus constructed versatile criteria,detecting genuine multipartite entanglement in a broad variety of cases.All famous examples of four-qubit multipartite entangled states are detected (GHZ,W ,Dicke,and singlet states)using the same criterion,without any optimization involved as the norms of any matricization of a correlation tensor are invariant under local unitary transformations on the density matrix (see Sec.VI ).Although for some specific states optimizing over all possible witnesses can yield a higher noise resistance in some cases,a comparable result is achieved in a computationally far more efficient way.As shown in Fig.2,there even exist states that were not detected to be genuinely multipartite entangled with any of the optimized criteria so far.V .DETECTION OF NON-FULLY-SEPARABLE STATESAs mentioned above,we can also use correlation tensorsto discriminate states containing some form of entanglement and fully separable states.This has already been carried out in [16],where the authors show that for fully separable states an upper bound on the trace norm of the full correlation tensor must hold for any matricization of the form one particle versus the rest (i.e.,matrix unfoldings).However,remarkably,our picture allows not only for a very simple proof of this fact,but,also,for a significantly stronger result since we can show that such a bound must hold for any possible matricization of the correlationtensor.FIG.2.(Color online)Here the parameter regions for whichthe state ρ=αρGHZ(2)+βρD (2)+1−α−β16I exhibits genuine multi-partite entanglement are identified.The GHZ and two-excitation Dicke state for four-qubit systems are defined as ρGHZ(2)=|GHZ(2) GHZ(2)|with |GHZ(2) :=1√2 1i =0|i,i,i and ρD (2)=|D 42 D 42|with |D 42 =1√6(|0011 +|0101 +|1001 +|1010 +|1100 +|0110 ).The (red)region labeled II uses criterion II from Ref.[23]optimized numerically over all local unitary representationsof the density matrix.The (yellow)region labeled I 42uses criterion I 42from Ref.[26]optimized numerically over all local unitary representations of the density matrix.The numerical optimization was performed using the composite parametrization from Ref.[38].The (blue)region labeled C shows the region detected to be genuinely multipartite entangled using Theorem 3.Theorem 4.For any fully separable state any matricization of the full correlation tensor must fulfill||T i 1···i k i k +1···i n ||trn j =1 2(d j −1)d j ,(23)i.e.,k =1,...,n −1and all possible permutations of theparticles are taken into account.Proof.Since the correlation tensor of a fully separable pure state must fully factorize into the one-body correlation tensors,it is straightforward to note that||T i 1···i k i k +1···i n ||tr= T (1)i 1···T (k )i k T (k +1)i k +1···T (n )i n tr=||(T (1)⊗···⊗T (k ))·(T (k +1)⊗···⊗T (n ))T ||tr =||T (1)⊗···⊗T (k )||||T (k +1)⊗···⊗T (n )||= j||T (j )||.(24)Using Eq.(3)the proof is finished.Notice that similar bounds hold as well for lower-order correlation tensors.Of course,one could also consider other。