Knowledge Representation,Sharing and
Retrieval on the Web
Philippe Martin and Peter Eklund
Distributed System Technology Centre,Australia
philippe.martin@https://www.doczj.com/doc/1c6516582.html,.au
Abstract.By“knowledge retrieval”,we refer to the automatic retrieval of statements permitting a tool to make logical inferences and answer queries precisely and cor-rectly,as opposed to retrieving documents or statements“related to”the queries. Given the ambiguity of natural language and our current inability to make computers “understand”it,the knowledge has to be manually encoded and structured using a formal graphic/textual language and ontologies(structured catalogs of categories and associated constraints of use).
The Web currently contains a lot of data,more and more structured data(databases, structured documents)and simple metadata but very little knowledge as de?ned above, i.e.very few knowledge representations.Moreover,this knowledge has been encoded using various languages and unconnected or loosely connected ontologies,and following di?erent representation conventions.Hence,currently,not only knowledge sources are rare but each require the development of a special wrapper for their knowledge to be interpreted and hence retrieved,combined or exploited.
This article reviews various projects concerning knowledge representation,sharing and retrieval on the Web,then details requirements for a“Semantic Web”and il-lustrates them with notations,conventions and cooperation rules from our own tool, WebKB-2.Knowledge retrieval mechanisms and interfaces used in WebKB-2are also given as illustrations.
Table of Content
1Introduction
2Elements and Landmarks of Knowledge Representation and Sharing on the Web 2.1Exchange Formats and Programming Interfaces
2.2Ontologies and Knowledge Bases
2.3Ontology Servers
2.3Knowledge Within Web Documents
3Requirements for a Viable Semantic Web
3.1Need for a Standard Library of Ontological Primitives
3.2Need for Expressive Notations
3.3Need for High-level(and Expressive)Notations
3.4Need for Lexical/Structural/Ontological Conventions
3.5Need for Flexible Ways to Refer to a Category
3.6Need for a Shared Natural Language Ontology
3.7Need for More Centralization
4Mechanisms for the Cooperatively Editing a Shared KB
5Search Interfaces and Mechanisms
5.1Searching Categories and Links
5.2Accessing or Adding Graphs Via Generated Interfaces
5.3Mechanisms for Searching Graphs
6Conclusion
1Introduction
Data indexed by formal terms(categories)or organized by inclusion links is called“structured data”.Beliefs,de?nitions,facts or rules represented using a formal language and an ontology(i.e.a catalog of concepts and relations,with constraints of use and organized by semantic links)is most often called“know-ledge”(or knowledge representations/statements).The Word Wide Web cur-rently contains a lot of data,more and more structured data(on-line databases, structured documents)and simple metadata but very little knowledge1.
Since the semantics of natural language sentences or structured data cannot be automatically extracted,people have to explicit semantics via knowledge representations.Then,tools may make logical inferences to answer queries pre-cisely and correctly(as opposed to retrieving data/documents“related to”the queries)and without requiring the users(or application developers)to know the exact schema(structure and/or indexation categories)used in each source of data.However,knowledge sharing,merging and retrieval are only possible if the categories used in the knowledge representations are connected by semantic links,directly(by belonging to a same ontology)or indirectly(by belonging to interconnected ontologies).
A common expectation for the“Semantic Web”2is that many small,spe-cialized(and possibly competing)RDF schemas(ontologies in RDF3)will be developed,and that in order to make knowledge representations,people or businesses will select some schemas,re-use them,and create new RDF schemas to de?ne terms they have not found[5].Then,according to Tim Berners-Lee4, future Web search engines may be able to?nd various statements related to certain queries and logically combine a few of them to answer the query;“while nothing will make the combinatorial explosion go away,many real life problems can be solved using just a few(say two)steps of inference out on the Web”.
This vision seems unrealistic since it is unlikely that di?erent people will create statements that can be logically matched or combined when they use unconnected or loosely connected ontologies,when only a few ontological prim-itives are standardized(in the RDF world,these are the categories of RDFS and DAML+OIL5),and when no lexical/ontological/structural/semantic con-ventions are adopted.Like today,future Web search engines would mostly rely on term lexical matching,and applications would have to write a special wrapper for each knowledge source they want to utilize(furthermore,even wrappers 1Although describing the same facts as we do,some researchers unfortunately use the words“knowledge”instead of“data”and“ontology”instead of“knowledge”, as in[3]:“The WWW can be viewed as the largest knowledge base that has ever existed.However,its support in query answering and automated inference is very limited”.The terms we use are chosen to be quite unambiguous for a member of the knowledge acquisition/representation community,and their meanings are consistent with the ones given in the“free on-line dictionary of computing”at https://www.doczj.com/doc/1c6516582.html,/foldoc/
2https://www.doczj.com/doc/1c6516582.html,/2001/sw/
3https://www.doczj.com/doc/1c6516582.html,/RDF/
4https://www.doczj.com/doc/1c6516582.html,/DesignIssues/Semantic.html
5https://www.doczj.com/doc/1c6516582.html,/2001/03/daml+oil-index
cannot compensate for badly structured or impoverished knowledge).Matching6 or combining statements,and hence?nding knowledge relevant to a query(or even only“related”to a particular object)is a problem even in a single large knowledge base(KB)such as CYC7where knowledge providers are trained knowledge engineers following conventions,using a unique large ontology and an expressive knowledge representation language.Yet,even in this ideal case, some choices in the ontological and structural conventions have led to knowledge which is not explicit enough to be exploited in many applications8.
The above cited vision also seems undesirable because,as we will show,more centralized approaches involving large-scale knowledge servers can(i)permit a large number of users(people or agents)to cooperatively build large knowledge bases(KBs)with explicit,expressive,normalized,highly inter-connected state-ments and categories,and hence permit knowledge retrieval by simple hypertext navigation or provide reasoning services at selected levels of e?ectiveness and completeness,and(ii)exploit the KB to ease,guide and cross-check the insertion of new knowledge by each user and her re-use/annotation/correction of other users’knowledge.These features are permitted by the incremental insertion of knowledge into centralized repositories when it is developed,instead of afterwards by Web search engines(knowledge in isolation is not knowledge but merely data; hence,loosely connected schemas/RDF documents cannot be logically combined and their re-use requires the development of an ad-hoc wrapper for each one). To keep the advantages of the decentralized approach of the Web,categories and statements in a knowledge server must be referable via URLs(and then exported in a standard language such as KIF),and be allowed to refer to other objects on the Web via URLs.These are easy-to-achieve constraints.
For e?ciency reasons,all Web-users cannot use the same knowledge server but they can use several general knowledge servers(e.g.managed by portal com-panies)and more specialized knowledge servers dedicated to speci?c domains. By(partly)mirroring one another’s content,general and specialized servers would share a similar general ontology like WordNet9or CYC’s ontology,and competing specialized knowledge servers would also share some similar content10. Thus,it would not matter where a Web user publishes information?rst,no unique server would have to be relied upon,and hence this“more centralized”approach would maintain the advantages of the current decentralized approach, 6For example,the statement“John is owner of a duplex in Southport”can easily be identi?ed as a specialization of the query statement“a person is owner of an apartment in a city part of the Gold Coast”provided that“John”has been declared as a“person”,“duplex”as a specialization of“apartment”and“Southport”as being a“city”which is“part”of the“Gold Coast”.
7https://www.doczj.com/doc/1c6516582.html,/tech.html#cycl
8For example,actions/processes are represented as n-ary relations instead of concept nodes with explicit thematic relations to the related objects.As shown in Section3.4, this decreases the possibility of matching or combining statements about processes. 9https://www.doczj.com/doc/1c6516582.html,/?wn/
10The similarity of the KBs also permits the processes of mirroring and answering queries involving several KBs.
without its problems.(A similar architecture for distributed KBs and a small-scale implementation is discussed in CYC11).
In this article,we?rst review some projects about knowledge representation, sharing and retrieval on the Web.Subsequently,we show that within a KB as well as across the Web,knowledge sharing and exchange implies that knowledge providers use of a unique set of ontological primitives,follow lexical/structural/se-mantic recommendations,and use(directly or via interfaces)high-level expres-sive knowledge representation languages that ease the adoption of the recom-mendations and lead to comparable12knowledge representations.Thus,we argue that the Semantic Web implies the standardization of such elements,and show the elements adopted and implemented in our public knowledge server and annotation tool,WebKB-213.We summarize the protocols used in WebKB-2 to permit the asynchronous cooperative building of the KB by the users,and ?nally present search interfaces and mechanisms.
2Elements and Landmarks of Knowledge Representation and Sharing on the Web
RDF was not the?rst step towards knowledge sharing on the Web,and in many aspects can be seen as a backward step.The list of elements and landmarks below is organised thematically but also chronologically.It is by no means exhaustive but the selected landmark tools or ontologies continue to be the most known or usable nowadays.Readers that are familiar with the?eld of knowledge sharing and the Semantic Web can restrict their attention to Subsection2.3only.
2.1Exchange Formats and Programming Interfaces
KIF(Knowledge Interchange Format)14is a low-level but expressive language (?rst order logic plus contexts and sets)originally designed in1992to per-mit translations between more specialized knowledge representation languages (KRLs).It has become the de-facto standard for expressing the semantics of KRLs in a computable way.It is complemented by the Ontolingua library15 which formalizes(in KIF)various elements necessary for expressing the seman-tics of KRLs,e.g.sets,relations,functions,numbers and frames.By comparison, 11https://www.doczj.com/doc/1c6516582.html,/applications.html#dai
12Category/statement comparability is an important notion in this article since know-ledge retrieval or inferencing is based on statement comparison and combination.
A statement(resp.a category)is comparable to another statement(resp.category)
if it generalizes it or specializes it.Logic generalizations are logic deductions.Some generalizations are simply structural simpli?cations and not logic deductions.This is detailed at the beginning of Section3.4.
13Usable at https://www.doczj.com/doc/1c6516582.html,
14https://www.doczj.com/doc/1c6516582.html,/kif/dpans.html
15https://www.doczj.com/doc/1c6516582.html,:5915/
RDF is also a low-level language but far less expressive,unsuited for logical inferences16and as an interlingua(this will be discussed in Section3).
KQML(Knowledge Query and Manipulation Language)17is a KIF-based message format and message-handling protocol designed in1994to support run-time knowledge sharing among agents.It has been re-used or extended by various other agent communication languages.
GFP(Generic Frame Protocol)18(1995)is a set of functions that supports a generic Application Programming Interface(API)for frame representation systems(FRSs).Various FRSs have implemented a GFP server,e.g.Loom19and SRI20.The GKB-Editor(Generic Knowledge Base Editor)21is a GFP client that permits the graphical browsing and editing of FRSs that have GFP servers.GFP was extended and replaced by OKBC(Open Knowledge Base Connectivity)22 in1998.
2.2Ontologies and Knowledge Bases
Ontologies are catalogs of categories with their associated complete or partial formal de?nitions which can also be seen as“constraints of use”.Complete de?nitions are de?nitions of necessary and su?cient conditions(to be instance of the category).Partial de?nitions may be prototypes(listing“probable”rela-tionships),de?nitions of su?cient conditions,de?nitions of necessary conditions (e.g.“subtype of”and“instance of”links from a category to another),etc. Categories may be relation types(called“properties”in RDF;they include functional relation types or“functions”),concept types(called“classes”in RDF) and individuals(class instances that are not classes themselves).
Some ontologies are about mathematical entities(e.g.sets,relations,func-tions,numbers,sequences and bags),or about relationships from/to physical dimensions(e.g.space,time and matter),or about a particular domain(e.g. elevators and chemical elements).They are often called“theories”,are generally small,and may include,generalize,specialize or compete with other theories. Since1993,the Ontolingua server23has hosted a library of such ontologies and permitted Web users to add new theories or combine theories to create knowledge bases.
Some ontologies classify all the concepts of a natural language or a particular domain,via links such as“subtype of”,“instance of”and“part of”.They are often called lexical ontologies and may be large.For example,WordNet24[11]is a“lexical database for English”that was Web-accessible as early as1990,and now connects about337,200words to about109,400concept types,and organizes 16For example,see www-rdf-logic mailing list archive at
https://www.doczj.com/doc/1c6516582.html,/Archives/Public/www-rdf-logic/
17https://www.doczj.com/doc/1c6516582.html,/kqml/
18https://www.doczj.com/doc/1c6516582.html,/?gfp/
19https://www.doczj.com/doc/1c6516582.html,/isd/LOOM/LOOM-HOME.html
20https://www.doczj.com/doc/1c6516582.html,/?sipe/
21https://www.doczj.com/doc/1c6516582.html,/?gkb/
22https://www.doczj.com/doc/1c6516582.html,/?okbc
23https://www.doczj.com/doc/1c6516582.html,:5915/
24https://www.doczj.com/doc/1c6516582.html,/?wn
these types via various kinds of links,e.g.specialization,exclusion,similar, member,part and substance.
Some ontologies classify relation types(e.g.spatial/temporal/thematic rela-tion types)and/or very general concept types(e.g.the notions of situation,state, process,spatial entity,physical entity)mainly via“subtype of”links.They are often called top-level ontologies.Examples are John Sowa’s ontologies25(1984 and2000)and the Generalized Upper Model26(1994).
Top-level ontologies may be used for structuring the lop layers of lexical ontologies.For example,Sensus[6]was created in1994by semi-automatically merging WordNet,LDOCE(the Longmann Dictionary of Contemporary En-glish)and two top-level ontologies:the Generalized Upper Model and Ontos. Similarly,in1995,we have used Sowa’s?rst top-level ontology to structure WordNet top layers and hence permit semantic checking on the use of WordNet categories[7].In1998,HPKB upper27was created by combining Sensus top-level ontology with CYC top-level ontology28.
Categories of lexical ontologies may be used as generalizations for the cate-gories in theories(which are generally much more precisely de?ned)and hence permit the retrieval and comparison of these categories and theories.This was also a goal for our work in1995.
A knowledge base(KB)is composed of one ontology(or several intercon-nected ontologies)plus additional statements using these ontologies.The classi-?cation of certain statements as being inside or outside the ontology is only an implementation dependent issue.Such a distinction does not need to be made in WebKB-2.When this distinction is made in other tools,statements that involve individuals and no universal quanti?er,are likely to be considered as not belonging to the ontologies.We do not make any distinction when we use the word“knowledge”.
2.3Ontology Servers
Ontology servers can also be called KB servers but the emphasis on the ontology highlights the fact that they permit Web users to modify the ontology part of the KB,which other KB servers do not allow(this technical limitation/simpli?cation is also why database servers do not allow the interactive modi?cation of the database schema).
The Ontolingua server29was probably the?rst ontology server(1993)and remains active.It permits the use of KIF?les or an HTML interface.The reading and editing of each“theory”may be restricted to a group of users but, apart from locking/session mechanisms,no particular support for synchronous or asynchronous cooperation between users is provided.
Ontosaurus30(1996)is also an ontology server with an HTML interface and permits each user to build or edit theories.It exploits the Loom31FRS.
25https://www.doczj.com/doc/1c6516582.html,/?sowa/ontology/
26http://www.darmstadt.gmd.de/publish/komet/gen-um/node1.html
27See HPKB-UPPER-LEVEL-LATEST in Ontolingua
28https://www.doczj.com/doc/1c6516582.html,/cyc-2-1/cover.html
29https://www.doczj.com/doc/1c6516582.html,:5915/
30https://www.doczj.com/doc/1c6516582.html,/isd/ontosaurus.html
31https://www.doczj.com/doc/1c6516582.html,/isd/LOOM/LOOM-HOME.html
The Co4system32(1996)permits some asynchronous cooperation between users via protocols modeled on submission procedures for academic journals, i.e.on peer-reviewing.The result is a hierarchy of KBs,the uppermost con-taining the most consensual knowledge while the lowermost KBs are the KBs of contributing users.This approach leverages some problems of interconnecting and comparing independently developed ontologies but doubtfully scales to large numbers of users.
Tadzebao and WebOnto33(1998)support some synchronous cooperation between co-temporal users(they can exchange multimedia messages and be warned of each other’s actions).
As opposed to most other ontology servers,WebKB-134[8](1998)does not store knowledge onto the server disk but can load and interpret Web-accessible?les that combine text,images and knowledge in various formats (mainly Conceptual Graphs[12]and Formalized English35).The knowledge parts are isolated via delimiters,e.g.the XHTML tags
WebKB-237[10](2001)inherits most of the features of WebKB-1(although its indexation and command languages are more limited)and also permits users to store knowledge into a unique KB on the server disk.As opposed to most other ontology servers,the knowledge from the various users is not stored into various loosely connected ontologies but tightly integrated into a same ontology/KB that has WordNet as backbone(thus,categories and statements are easier to retrieve,compare and re-use).Lexical problems are avoided by pre?xing each category identi?er with the identi?er of its source(user,organization,document, ontology,...).Lexical facilities are provided by the distinction between category identi?ers(which are unique)and category names(which may be shared).The cooperative building of the KB is supported via the enforcement of editing rules (presented in Section4).This type of asynchronous cooperation is likely to be more scalable in the numbers of users and knowledge quantity than Co4’s and leads to a better integration of the knowledge.WebKB-2also departs from other ontology servers by the size of the ontology it can manage.At present,WebKB-2 integrates various top-level ontologies plus the part of WordNet1.7concerning nouns(i.e.108,000nouns connected to74,500categories organized by various 32http://ksi.cpsc.ucalgary.ca/KAW/KAW96/euzenat/euzenat96b.htm
33http://ksi.cpsc.ucalgary.ca:80/KAW/KAW98/domingue/
https://www.doczj.com/doc/1c6516582.html,
35https://www.doczj.com/doc/1c6516582.html,/doc/languages/
36https://www.doczj.com/doc/1c6516582.html,/kb/sisyphus1.html
https://www.doczj.com/doc/1c6516582.html,
kinds of links).One of the few other KB systems that can manage large and dynamically modi?able ontologies is Parka-DB system38(1994).
Many RDF parsers exist but we do not know of any ontology server capable of really exploiting RDF,although Corese39does convert some RDF input into simple Conceptual Graphs40(CGs)and then is able to retrieve them by looking for specializations of a query graph.RDF export is simpler to implement than RDF exploitation,but RDF exports are either restricted to very simple know-ledge or are ad-hoc,needs the use of extensions,and therefore requires a special interpretation to be re-used.Ontobroker(discussed in the next subsection)does some exports in RDF.WebKB-2can export links between categories in RDF. In[9](2000),we proposed conventions and extensions to RDF/XML for the represention of various knowledge representation cases related to the use of contexts,negation,universal quanti?cation,collections,intervals,declarations and de?https://www.doczj.com/doc/1c6516582.html,te2001,we extended this work,taken into account the recent developments of the DAML+OIL top-level ontology,and showed how each of these cases could be represented in Formalized English,KIF and Conceptual Graphs and,to a certain extent,RDF/XML41.We have begun to implement the import and export of RDF/XML in WebKB-2with respect to these conventions and extensions.
2.4Knowledge Within Web Documents
SHOE42(1996)was the?rst(well known)language and server permitting the in-sertion of knowledge within HTML documents.Its claims,approach and notation were(and still are)surprisingly similar to the views and reasons of the W3C for the“Semantic Web”43(1998)and its recommended notation,RDF/XML.XML was?rst published as a W3C recommendation in1998.The RDF model and its XML notation,RDF/XML,were designed to permit the representation and sharing of knowledge(which is di?cult to do directly with XML)44.They were published as a W3C recommendation in1999.In2000,the syntax of SHOE was slightly modi?ed to be XML-compliant.Unlike RDF/XML documents,SHOE documents can be any HTML document into which some SHOE markups have been added.However,since it is XML-based,that knowledge is di?cult to read and write by people.Furthermore,the knowledge within a document is restricted to be about the object represented by the document(e.g.if a?le is about a certain person,its URL becomes an identi?er for that person and the knowledge lists relationships from that person to other objects).Hence,it seems there must be as many documents as individuals.Furthermore,categories cannot be declared/de?ned and used in the same document.Many knowledge-oriented 38https://www.doczj.com/doc/1c6516582.html,/projects/plus/Parka/parka-db.html
39http://www-sop.inria.fr/acacia/soft/corese.html
40https://www.doczj.com/doc/1c6516582.html,/?delugach/CG/
41See https://www.doczj.com/doc/1c6516582.html,/doc/translations.html
42https://www.doczj.com/doc/1c6516582.html,/projects/plus/SHOE/
43https://www.doczj.com/doc/1c6516582.html,/DesignIssues/Semantic.html
44https://www.doczj.com/doc/1c6516582.html,/DesignIssues/RDF-XML.html
XML-based notations(with associated parsers and sometimes inference engines) now exist,e.g.Rule-ML45,OML46,DAML47and RDF/XML.
Noticing that XML-like notations force the author of a text document to duplicate the information into a di?cult to write formalized version,the authors of Ontobroker[3]48(1998to2000)have opted for a tight integration to HTML: instead of introducing new tags,they ask the document authors to insert an attribute called”onto”into anchor tags and use the value to formalize the destination of the anchor.For example,
means that Richard(identi?ed by his home page URL)is a researcher.If this metadata is in Richard’s home page,it may be abbreviated:. Under the same conditions,
IIIA means that Richard’s a?liation is http://www.iiia.csic.es/.Each document had to be registered to the Ontobroker server(which accessed the registered docu-ments from time to time).The ontology could not be de?ne in the document but had to be de?ned in the Ontobroker server.More precisely,modi?cations to the ontology had to be submitted to the person authorized to modify the ontology and discussed by the people using it.To our knowledge,the ontology within On-tobroker’s web-accessible server was very small(a few dozens categories mainly about research domains and researcher/student levels).Hence,the users could simply index their research domains and professional status with the available categories.They could not actually represent the content of their research,or anything else,since they could not declare new categories.Because of these restrictions and the poor expressivity of the KRL illustrated above,Ontobroker (claimed by its authors to be the“?rst Semantic Web server”)was mainly used as a small database server.This is not to say that Ontobroker could not have been used as a normal ontology server since it could also parse Frame-Logics,a ?rst-order logic frame-oriented language,but it seems that this language could only be used as a query language by Web users.To sum up,the way Ontobroker was proposed to Web users was probably the most unscalable and unusable way imaginable.
As introduced above,WebKB-1(1998)and WebKB-2(2001)exploit a third way to store knowledge in HTML documents:high-level expressive and easily readable knowledge representations that are separated from the rest of the docu-ment by special delimiters.(Examples are given in Section3.2).If necessary,the representations may also be hidden by enclosing them between HTML comment tags.The user may mix category declarations/de?nitions and other statements, as long as a category is declared before being used.With WebKB-2,the user may also re-use the categories of the shared KB(about77,000at the beginning of 2002)via their identi?ers or,when there is no ambiguity,via one of their names. Then,when satis?ed with the content of the document,the user may commit it to the KB(even with high-level languages,knowledge modelling is not unlike 45http://www.dfki.uni-kl.de/ruleml/
46https://www.doczj.com/doc/1c6516582.html,/OML/OML-Examples.html
47https://www.doczj.com/doc/1c6516582.html,/
48https://www.doczj.com/doc/1c6516582.html,/
programming:it requires various syntactic and semantic checking,corrections and sometimes re-organizations).
HTML hyperlinks and some other HTML tags,especially the de?nition tag, can be viewed as a high-level,easily readable but poorly expressive way of encoding knowledge.For example,in WebKB-1,the following statements in Formalized English(FE),Frame-CG(FCG)and HTML were equivalent:
FE:The car that has for owner John has for weight1750kg.
FCG:[the car,owner:John,weight:1750kg]
HTML:
3Requirements for a Viable Semantic Web
As highlighted in the introduction,we think the vision of the Semantic Web as a collection of documents that re-use barely connected RDF schemas is not only unrealistic but undesirable.In this section,we list some elements that are necessary for the realization of the Semantic Web.
3.1Need for a Standard Library of Ontological Primitives
RDF is not a particularly expressive language even with the semantic augmenta-tions provided by the“standard”schemas RDFS and DAML+OIL.For instance, we have not found any(non ad-hoc)way to represent simple sentences like “5persons dance together”50or“51%of people are women”in RDF.Many logic-related problems with RDF can be found in the www-rdf-logic mailing list archive51.
The lack of expressiveness of RDF and the absence of standard ontological primitives force knowledge providers to represent information in a biased or impoverished way or invent their own(mutually incompatible)extensions.Both cases make knowledge exploitation,sharing and re-use di?cult.
Many formal speci?cation languages such as Z52come with a mathematical toolkit,i.e.functions and relations related to the building blocks for knowledge 49https://www.doczj.com/doc/1c6516582.html,/2000/07/hs78/
50There is no“set”class nor“size”property/relation in RDF,RDFS or DAML+OIL.
There is a“cardinality”property in DAML+OIL but it is about the number of relations that instances of a certain class can have.Representing“together”is also
a problem since there is neither a way to represent that an universally quanti?ed
variable is within the scope of an existentially quanti?ed variable,nor a special keyword to specify a“collective”interpretation for a collection(RDF only proposes the“distributive”and“cumulative”interpretations).
51https://www.doczj.com/doc/1c6516582.html,/Archives/Public/www-rdf-logic/
52https://www.doczj.com/doc/1c6516582.html,/?mike/zrm/
representation:sets,relations,functions,numbers,sequences and bags.KIF,the best accepted knowledge exchange format,also comes with a similar toolkit and is complemented by the Ontolingua library.
A similar mathematical toolkit needs to be standardized in schemas such as RDFS to permit knowledge representation,sharing and exploitation.For example,RDF engines cannot provide an implementation for handling sets, general negation or universal quanti?cation if a vocabulary is not?xed.(We recognize the DAML+OIL schema is a?rst step in that direction).
3.2Need for Expressive Notations
A usual concern about expressive notations is that they are too complex to handle e?ciently.Actually,it is more correct to say that when statements use complex features and when these complex features are exploited by an inference engine for logical inferences,this inferencing may not be e?cient.However, when statements are biased because the notation is too restrictive or the user is not precise,they cannot be exploited for(correct)logical inferencing by any application.
Most KRLs are customized for a particular inference engine and are not expressive enough for precise representations of natural language sentences,and hence,information in general.However,information or knowledge on the Web are not dedicated to a particular application.A restricted model and notation such as RDF+RDFS+DAML+OIL and RDF/XML cannot be used as an inter-lingua because it arbitrarily limits the expression and exploitation of knowledge representations.
On the other hand,there is no harm in using expressive notations since,an inference engine is not obliged to take account of all the features of the language (i.e.all the categories in the“standard”schemas/ontologies)and perform all the logical deductions.What inferencing is done is an application-dependant choice, and not simply for e?ciency reasons:the kinds of rules to apply(e.g.to handle modalities)can also sometimes only be chosen according to the application. Hence,the issues of completeness and decidability are not related to notations but to inference engines.
In the HTML/XML worlds,applications ignoring parts of the structured data is a tradition.In the speci?cation of some knowledge exchange languages and APIs,such as KIF and OKBC,various levels of conformance for compliant inference engines are listed.Alternatively,each inference engine may advertize the categories to which they accord a special interpretation(and thence which features they exploit and how).
Inference engines do not even have to exploit category de?nitions:they may implement some e?cient ad-hoc exploitation of them(the formal de?nitions still permit the semantics of the categories to be speci?ed and permit the programmer to know and delimit the kinds of deduction the implementation performs). Techniques to search specializations of a query graph[12],or more generally,path retrieval techniques(based on structural matching and exploiting specialization links between categories),can be e?cient53and provide satisfying results for 53If the query graph and each of the statements has a tree structure,the search complexity is polynomial(see[1]).
knowledge retrieval.For example,by treating a relation/property“not”as if it had no special meaning,WebKB-2can e?ciently retrieve the representations of “there is no duplex for rent in Southport”and“Southport is part of the Gold Coast”in answer to the(formalization of the)query“Is there an apartment for rent on the Gold Coast?”.In this example,as opposed to the one given in Footnote6,the results are not“logic specializations”of the query but nonetheless relevant answers.More details will be given in Section3.4and Section5.3.
3.3Need for High-level(and Expressive)Notations
A problem for automatic knowledge retrieval and inferencing is that a same piece of information can be expressed in many di?erent incomparable ways. This problem is particularly acute when a low-level general syntax such as KIF (LISP)or RDF(XML)is employed,or when standard schemas o?er partially redundant ontological primitives54.
Some ways to represent information are more explicit,re-usable,comparable and easier-to-handle than other ones.Hence,to improve knowledge use and re-use possibilities:(i)knowledge representation conventions(or“recommenda-tions”)should be standardized;(ii)high-level languages(or graphical interfaces) should guide the user and lead her to use the adopted conventions.We have proposed a minimal set of lexical/structural/ontological recommendations in[9]. We give a summary of these in the next section.These recommendations are also usefully observed within a KB server.WebKB-2users are asked to follow them and the high-level notations that we have designed–Frame-CG and Formalized English–encourage their adoption.
Frame-CG(FCG)55is a notation that we have derived from CGLF(the Con-ceptual Graph56linear form)[12]to improve on its readability and expressivity (which were already the main reasons for the success of Conceptual Graphs).The three main improvements were:(i)the introduction of many kinds of quanti?ers in the form of English articles or expressions(e.g.“many”,“between2and5”,“at least6.5%”);(ii)a shorter and more natural way to express relations between objects;and(iii)the convention that the scope and precedence of quanti?ers in a graph(seen as a logic formula)are related to the graph structure and node order(as in predicate logic)57.
Formalized English(FE)is identical to FCG apart from some syntactic sugar used for grouping and connecting objects.The model behind these notations(i.e. the model implemented in WebKB-2)58may be seen as a generalization of the 54For example,to represent an“xor”between two statements,one could think of using an RDF“alt”container,a DAML+OIL“disjointWith”relation(by creating an anonymous class for each of the statements)or a classic“xor”relation(e.g.KIF “xor”relation).
55Grammar in https://www.doczj.com/doc/1c6516582.html,/doc/F languages.html#FCG
56https://www.doczj.com/doc/1c6516582.html,/?delugach/CG/
57In CGLF,only contexts are important to determine the scope of quanti?ers;
otherwise,universal quanti?ers are assumed to have wider scope than existential quanti?ers except when the keyword“@certain”is associated to them;this convention leaves room to ambiguities.
58https://www.doczj.com/doc/1c6516582.html,/doc/dataModel.html
Conceptual Graph model,RDF and terminological logics.Like these models,it is a logic-based semantic network model and permits to store logical statements. The KIF model is not yet completely included but soon will be(we currently have some problems with sets and second-order statements).
To illustrate FCG and FE and compare them to the other cited languages, here is the representation of an English sentence in CGLF,FCG,FE,KIF, predicate logic(PL)and RDF/XML(the XML format for the RDF data model). Namespaces are omitted.“Ned”is assumed to be a declared identi?er for an instance of the type“Person”.The‘s’at the end of“cars”and“sells”in the FCG and FE representations are automatically removed by WebKB-2(since a universal-like quanti?er is used with these categories).
E59:Ned sold(the same)3cars twice on the21/1/2001.
CGLF:[Person:Ned]←(agent)←[Sell:{*}@2]-
{→(object)→[Car:{*}@3@certain];
→(time)→[Date:#21/1/2001];}
FCG:[3cars,object of:(2sells,agent:Ned,time:21/1/2001)]
FE:3cars are object of2sells with agent Ned and time21/1/2001.
KIF60:(forAllN3’?c car(forAllN2’?s sell
(and(agent’?s Ned)(object’?s’?c)(time’?s’21/1/2001)))) PL:?cars set(cars)∧size(cars,3)∧?c∈cars
?sells set(sells)∧size(sells,2)∧?s∈sells
agent(s,Ned)∧object(s,c)∧time(s,21/1/2001)
RDF61:
More translation examples can be found on the WebKB-2site62.
The need for higher-level(and more expressive)notations than RDF/XML is well recognized63.As“an academic excercise”,Tim Berners-Lee has begun the design of Notation364,another notation for RDF which has some points 59This sentence does not specify whether the cars have been sold individually,2by2, or3by3.This ambiguity is kept in the representations.
60Here is our KIF de?nition for the“forAllN”quanti?er:
(defrelation forAllN(?num?var?type?predicate):=
(exists((?s set))(and(size?s?num)
(truth?(forall(,?var)(?(member,?var,?s)(and(,?type,?var),?predicate))))))) 61This RDF representation is only a tentative.
62E.g.at https://www.doczj.com/doc/1c6516582.html,/doc/translations.html and
https://www.doczj.com/doc/1c6516582.html,/kb2/translation.html
63https://www.doczj.com/doc/1c6516582.html,/DesignIssues/Logic.html
64https://www.doczj.com/doc/1c6516582.html,/DesignIssues/Notation3.html
in common with CGLF,FCG,FE and frame languages.(However,Notation3 does not(yet)have any special syntax for extended quanti?ers,collections, functions and de?nitions).Although Berners-Lee writes that he has not designed Notation3“as an alternative to RDF’s XML syntax which has the fundamental advantage that it is in XML”,one may wonder what this advantage is supposed to be since he also acknowledges that most notations may be“Web-ized”65by using URIs for category identi?ers.Even if knowledge can be represented in XML,it is unlikely that XML objects are directly used by advanced inference engines,and that knowledge providers read or write XML-based languages. Hence,translations to and from the XML world are necessary.From a purely syntactical viewpoint,the use of a Lisp-like notation(such as KIF)as a general low-level interlingua makes more sense because Lisp is concise and has adequate quotation(contextualization)features.
From any viewpoint we can think of,the use(and ideally,the standardi-zation)of a high-level expressive notation would make even more sense since then knowledge is easier to write,read,compare,exchange and exploit66.Being readable and not XML-based,knowledge representations can also be mixed and hyperlinked with text and images within HTML/XML documents(WebKB-1 and WebKB-2exploit such documents).
3.4Need for Lexical/Structural/Ontological Conventions
Consider the statements“a person is doing something”and“Ned is selling a car”and their FCG representations[a person,agent of:an activity]and [Ned,agent of:(a sell,object:a car)].The second graph is a speciali-zation of the?rst,i.e.it has more information in its structure(one more relation) and in its components(“Ned”is an instance of the type“person”and“sell”is a subtype of“activity”).Therefore,since only existential quanti?ers are involved in those graphs,the second logically entails the?rst67.In other words,if the ?rst is used as a query graph,the second is a logical answer.
Similarly,the second graph can also be seen as a specialization of the FCG given in the previous example but,since it involves universal quanti?ers,there is no logical entailment relation between the two graphs.Hence,we simply say the graphs are comparable(in the same way that two categories are comparable if they are linked by a subtype link or an instance link).
Now,suppose that a user declares a relation type(“property”in RDF)“sell”to represent the information“A person sells a car”via2nodes linked by a relation;in FCG:[a person,sell:a car].This graph leaves the“agent”and “object”relations implicit and is not comparable to any of the previous graphs. The user could associate a de?nition to the relation type“sell”to permit the expansion of the previous graph to:[a person,agent of:(a sell,object: 65https://www.doczj.com/doc/1c6516582.html,/DesignIssues/RDFnot.html
66Let us stress again that a high-level expressive language such as FCG or FE is not intended to limit what the knowledge provider can express but how she express it, and furthermore its expressiveness does not impose constraints on what inference engines must do.
67For more details and a mathematical proof,see[1].
a car)]but such an expansion can be a complex process and few inference engines perform it.The relation type“sell”cannot be re-used when other rela-tionships(such as“time”or“purpose”)have to be represented,and would be incomparable with other relation types“sell2”and“sell3”used to represent these relationships.Furthermore,relations cannot be quanti?ed.In summary,the use of relations other than basic binary relations should be avoided because this use leads to representations that are less explicit and comparable.Even if a Web-based knowledge-oriented information retrieval engine does some lexical match-ing on category names to complement structural/semantic matching,concept types“sell”are more likely to be used in unrelated KBs(if basic binary relations are used)than relation types such as“sell2”or“sellSomethingAtSomeTime”(these kinds of identi?ers are quite typical when relational/functional syntaxes such as Lisp are used).
As opposed to concept types,there is not a great number of basic binary relation types needed to represent natural language.For example,WebKB-2has about74,500concept types derived from the WordNet lexical database about nouns,but it has a stable ontology of only140relation types and50of these types appeared su?cient to us for representing most usual natural languages sentences.Basic binary relation types are an e?cient way to guide and normalize the knowledge representation task.Thanks to the signatures associated with these relation types,an inference engine can easily perform some elementary semantic checking and propose corrections when signatures are violated.
Because of its Lisp-like syntax,KIF does not encourage the use of basic bi-nary relations only.Like most frame-based or graph-based languages,RDF only accepts binary relations but its cumbersome XML syntax discourages knowledge providers to be precise.For the same reasons,KIF and RDF discourage the use of adequate quanti?cation,and do not prevent the use of verbs,adverbs,and adjectives as category identi?ers/names even though such categories cannot be quanti?ed(e.g.“any qualify”and“3quali?ed”are meaningless),can rarely be compared to other categories,and leave information implicit.Thus,to permit knowledge sharing,lexical/structural/ontological conventions are required,and their observance needs to be encouraged by high-level notations.
RDF/RDFS and the“Meta Content Framework Using XML”68have some “naming conventions”for category identi?ers:words used should be singular, with a lowercase?rst letter for relation types and an uppercase?rst letter for other kinds of categories,and the intercap style should be adopted when the identi?er is composed of several https://www.doczj.com/doc/1c6516582.html,ing names in the singular is a sound convention because categories can then be quanti?ed in various ways(whereas for example a category“cars”cannot be used in a universally quanti?ed node and is not comparable to“car”).However,with the intercap style and the?rst letter in uppercase,the correct cases in the names may be lost and,at least in English, there is no way to recover that information.Readable and correctly spelled category identi?ers are needed when using the identi?ers in menus or presenting information with languages such as Formalized English(FE).(In RDF,correct spellings can be seci?ed via the label relation but this is a cumbersome and rarely used feature).
68https://www.doczj.com/doc/1c6516582.html,/TR/NOTE-MCF-XML/#secA.
Hence,a summary of a minimal set of conventions that we advocate is:
–lexical conventions.Whenever possible,use a correctly written English sin-gular noun or nominal expression for each category identi?er.Separation between words is to be done with underscores(dashes and quotes may be used when part of the usual spelling of words,e.g.“Niemann-Pick disease”
and“Fallot’s tetralogy”).
–structural/ontological conventions.Only use basic binary relations and res-pect reading conventions69.Whenever possible,use or specialize categories from standard ontologies and use the least expressive ontological primitives: try to avoid general negation,disjunctions,second-order statements,collec-tions,etc.In the RDF context,this amounts to using RDF and DAML+OIL ontological primitives whenever possible.Within WebKB-2,this amounts to selecting categories of the shared ontology and then following the menus or using the“For Ontology”(FO)notation for links between categories and FCG or FE for other kinds of statements.Via these notations and the ontology(relation types,general schemas/templates,etc.),the WebKB-2 user is guided to represent most things in a normalized way:states,pro-cesses,descriptions,indexations,characteristics,measures,numbers,collec-tions,temporal/spatial/logical entities/relations,etc.
–semantic conventions.Be as precise as possible:give adequate quanti?ers, contextualize statements in time,space,authorship,etc.Re-use and com-plement existing knowledge.To enforce this in WebKB-2,(i)a category can only be declared by connecting it with another and it must have at least one generalization or specialization,and(ii)a statement cannot be entered if it contradicts or is directly comparable to an existing statement,unless the author asserts the relationships between the two statements.
3.5Need for Flexible Ways to Refer to a Category
There are more e?cient and elegant approaches than others to avoid lexical problems in a KB even though there are multiple knowledge providers and multiple names for each category.
In RDF,a category is uniquely identi?ed by a URI,e.g.https://www.doczj.com/doc/1c6516582.html, and https://www.doczj.com/doc/1c6516582.html,/doc.html#car.Within a multi-user KB server,it makes more sense to use user identi?ers than document URIs as knowledge source identi?ers.Thus,in WebKB-2,a category identi?er can be a URI(or an e-mail address)but also the concatenation of the knowledge provider’s identi?er and a key name,e.g.wn#dog,wn#time,pm#IR_system(“wn”refers to Word-Net1.7and“pm”is the login name of the user represented by the category 69Most semantic networks models(including RDF)have adopted the convention that
a relation“R”from a node“A”to a node“B”should be read“the R of A is B”or
“A has for R B”.In models where relations can be of any arity(e.g.KIF)no such convention is generally advocated,resulting to various usages and interpretation problems(e.g.in the ontologies of KIF,the relations“subset”and“member”
counter-intuitively have the source set as a second argument instead of as the?rst).
philippe.martin@https://www.doczj.com/doc/1c6516582.html,.au).In this third case,the category may still be referenced from outside the KB by pre?xing the identi?er with the URL of the KB,e.g.https://www.doczj.com/doc/1c6516582.html,/kb/wn#time.This method is used when knowledge is exported in RDF/XML.
In addition to an identi?er,a category may have various names(which may also be names of other categories).In FE,FCG and FO,a category identi?er may show several names,e.g.wn#dog__domestic_dog__Canis_familiaris(at least two underscores must be used for separating the names).Given95%of current categories in WebKB-2come from WordNet,the“wn”pre?x may be left implicit, e.g.#time means wn#time.More precisely,“wn”is the default creator.An or-dered list of default creators can be speci?ed,e.g.“default creators:pm wn;”.
Below is the way the FO notation can be used in WebKB-2to store that the uppermost concept type has been created on the29/11/1999,given two names by its creator“pm”,that the user“oc”has added a French name and an“in-stanceOf”link to the RDF“class”category,that“pm”has added a disjointWith link to the uppermost relation type(the link creator is left implicit since it is the same as creator of the source category)and given three subtypes,two of which forming a close partition(or“disjoint union”in DAML terminology).
pm#thing__top_concept_type(^thing that is not a relation^)29/11/1999
_chose(oc fr),
^rdfs#class(oc),
!pm#relation,
>{(pm#situation pm#entity)}pm#thing_playing_some_role;
Here is a partial translation in RDF/XML.The creators of the links could not be represented in a standard/simple way.
xmlns:rdfs="https://www.doczj.com/doc/1c6516582.html,/TR/1999/PR-rdf-schema-19990303#" xmlns:daml="https://www.doczj.com/doc/1c6516582.html,/2000/10/daml-ont#" xmlns:pm="https://www.doczj.com/doc/1c6516582.html,/kb/theKB_terms.rdf/pm#"> rdf:resource="https://www.doczj.com/doc/1c6516582.html,/TR/1999/PR-rdf-schema-19990303#Class"/> rdf:resource="https://www.doczj.com/doc/1c6516582.html,/kb/theKB_terms.rdf/pm#relation"/>
Here is how the FO notation was used by“pm”to declare a category for the“instanceOf”relation,specify the equivalent RDF category and an inverse relation.
pm#kind__type__class(pm#thing,rdfs#class)
=rdf#type,
-pm#instance; Here is a partial translation in RDF/XML using the previous namespaces. rdf:resource="https://www.doczj.com/doc/1c6516582.html,/TR/1999/PR-rdf-schema-19990303#Class"/> rdf:resource="https://www.doczj.com/doc/1c6516582.html,/1999/02/22-rdf-syntax-ns#Type"/> rdf:resource="https://www.doczj.com/doc/1c6516582.html,/metadata/dublin_core#type"/> rdf:resource="https://www.doczj.com/doc/1c6516582.html,/kb/theKB_terms.rdf/pm#instance"/> More details on our top-level ontology and how it integrates other top-level ontologies can be found on the WebKB-2site(https://www.doczj.com/doc/1c6516582.html,). WebKB-2maintains links between the links between each category and its creator and names,and conversely.This permits the use of names instead of identi?ers within statements as long as there is no ambiguity.Relation signatures are exploited to eliminate candidate categories70.If there is more than one candidate for a category,the parsing stops and the list of candidates is printed to help the user re?ne her statement.For a query graph,there is no harm in making this choice automatically and let the user re?ne the query if an incorrect category has been selected.For improved readability,we often use names instead of category identi?ers in the example graphs of this article. A problem that prevents this facility to be adopted within RDF documents on the Web is that the RDF schemas they import may change(new names may be added to categories)and hence ambiguities may appear. Within a KB that integrates a natural language ontology,this facility is particularly useful to accelerate the writing of knowledge. 70For example,“?ight”is a name currently shared by9categories:4representing processes,3representing collections,1representing a psychological feature,and 1representing a physical entity(“?ight of stairs”).If a concept node is about a “?ight”and is the destination of a relation with type pm#on location,given the signature associated to pm#on location,only one sense of“?ight”is relevant,the one representing the physical entity“?ight of stairs”. 3.6Need for a Shared Natural Language Ontology Links from a natural language ontology such as WordNet71form the backbone of a large shared KB,and are a way to connect ontologies on the the Web. Such links permit WebKB-2to relate,compare and retrieve knowledge repre-sentations.They also provide the user with various categories(meanings)for a word,and various distinctions for a notion,many of which she may not have considered.This leads the user to enter more precise and comparable represen-tations.The semantic constraints associated with the top level categories of the ontology are inherited by all the categories of the natural language ontology,and this permits some automatic checking on all users’statements and extensions to the ontology. We initialized the current KB of WebKB-2with the content of the lexical database WordNet1.7:108,000nouns and74,500categories referred by nouns (in accordance with our lexical conventions,we ignored information regarding verbs,adverbs and adjectives). Various kinds of links connect these categories:specialization,exclusion, similar,member,part,substance,and their inverse links.The interpretation of links other than specialization,exclusion and similar are not always clear nor consistent within Wordnet.For example,a part link from the category airplane to the category wing could mean that“any airplane has for part at least1wing”or“all airplanes have for part the same wing”,“any wing is part of a plane”,etc.We assumed the?rst interpretation was correct for direct links(e.g.part,substance,etc.)and therefore opposite for their inverse links (part of,substance of,etc.).This interpretation is exploited in our graph comparison/retrieval algorithms. To permit the use of WordNet in a KB server,we have(i)generated a unique identi?er for each category,using the most commonly used word for that cate-gory,whenever it was possible;(ii)distinguished the Wordnet specialization links into subtype links and instance links by isolating about6000individuals, and(ii)re-structured and complemented WordNet top-level ontology with about 150concept types,and200relation types.Thanks to this re-organization and our ontology checking mechanisms,we detected about300semantic errors(e.g.ca-tegories specializing exclusive categories,redundancies,subsumption links used instead of part-of links or member-of links,etc.)and manually corrected them. We also made about500lexical corrections.(See https://www.doczj.com/doc/1c6516582.html,/doc/wn/for details). WordNet is also used in other knowledge-based systems,e.g.AI-trader72, a knowledge base broker,and Ontoseek[4],a knowledge retrieval system.Both permit their users to enter“simple”CGs(i.e.existentially quanti?ed and without contexts)for representing knowledge,and permit“queries for specializations of a query graph”for retrieving knowledge. 71https://www.doczj.com/doc/1c6516582.html,/?wn/ 72https://www.doczj.com/doc/1c6516582.html,rmatik.uni-frankfurt.de/projects/aitrader/intro.html 3.7Need for More Centralization According to Tim Berners-Lee73,“many KR systems had a problem merging or interrelating two separate knowledge bases,as the model was that any concept had one and only one place in a tree of knowledge...The RDF world,by contrast is designed for this in mind,...”.Although RDF schemas may indeed import other RDF schemas and RDF documents may import various RDF schemas,in order to compare two statements from di?erent RDF documents, an RDF engine has to classify the categories used in these statements into a unique specialization hierarchy74.This is most often impossible(unless the two documents mostly re-use the same schemas)because of the disconnected specia-lization hierarchies(and hence insu?cient information to compare the categories and statements).What are currently called“ontology-merging techniques”,are only semi-automatic algorithms heuristically matching categories based on their names,links to other categories75,and sometimes other properties such as their frequency of occurrence in documents[13]. From the knowledge provider’s view,re-using distributed RDF schemas is also a di?cult and sub-optimal task.First,she must?nd schemas on the Web with categories similar to the ones she wants to use,then select some schemas that are not mutually inconsistent and write another schema to de?ne the categories she has not found.Tools exploiting distributed schemas cannot provide guidance nor much cross-checking since they do not have a large ontology to exploit.In WebKB-2,thanks to the initialization of the KB with WordNet,the user enters a word and is presented with the categories that represent its various meanings,generalizations and specializations.She can select one category or ?nd a more appropriate category by navigating along semantic links.When a new category is required,the user can add it by connecting the new category to an existing category via a link of a selected type.Since the new category is added to a large and tightly interconnected ontology,it can be accessed and exploited in many ways.With distributed schemas,to achieve a similar level of connectedness,each schema creator would have to check that there is a relation between each of her categories and all relevant categories in all other existing schemas on the Web. There is an intermediate way between the highly decentralized approach advocated by the W3C and the approach we have adopted.That is to develop RDF schemas/documents by re-using(importing)ontologies of large KB servers such as WebKB-2.Then,tools could provide some guidance and cross-checking, and do a relatively good job at integrating these schemas/documents even when developed separately since they would at least be based on the same large natural language ontology.WebKB-2permits its categories or parts of its ontology to be referred and accessed via URLs and can import knowledge from Web documents into its shared KB,permanently or for testing puposes.However,the constraints are the same as when knowledge is entered manually,and an import is rejected if a problem is encountered.With the intermediate way,future Web search engines will have to be more permissive. 73https://www.doczj.com/doc/1c6516582.html,/DesignIssues/RDFnot.html 74Not simply a tree since a category may have several parents. 75See Chimaera at https://www.doczj.com/doc/1c6516582.html,/software/chimaera/ 一、阅读理解 1．课外阅读轩。 哑巴渡 离家不远的地方有一个渡口，摆渡的人是一个（）的哑巴老头，乡亲们都管这儿叫哑巴渡。 哑巴老人的船总是干净而清爽，船舷擦得亮亮的，闪着桐油的光泽，一点也不像他（）的脸。哑巴老人的船摆得又快又稳，无论是白天还是黑夜，只要在岸上一吆喝，他就会从船舱里出来，再（）地把过河人送到对岸去，春夏秋冬，从不间断。在这里过河的人，不用担心会耽搁行程。 河水在日复一日，年复一年地流淌着，日子在（）中逝去。哑巴老人在这里默默地把船摆过来，渡过去，人们已经习惯了河水、渡船和哑巴老人的存在，好像这一切都是理所当然的。 直到有一天，哑巴老人生病被送进了医院，大家才怀念起以前只用一双粗大的手说话的哑巴老人。人们想起他摆船的稳当。想起了他一毛钱的渡河钱。想起了他冬天半夜里穿衣解缆就为了送一个行人过河。哑巴不在了，人们觉得船上空落落的。 一天，乡里突然来了个老将军，手里捧着哑巴老人的骨灰盒。哑巴老人得的是肝癌，死在医院里。他把一生摆渡的钱全捐了出来，想在这条河上建一座新桥。 就在新桥建成剪彩的那天，老将军含泪讲述了一件让人震惊的往事：在红军长征的一次战斗中，为了拦截追击的敌人，还是“红小鬼”的老将军奉上级的命令配合班长炸掉石桥。摆渡的哑巴老人就是我的班长！ 阳光下，洁白的桥身闪烁着三个大字：哑巴渡。一位老红军生前在这里炸掉一座桥，摆了一辈子渡，死后又留下了一座桥。 （1）选择合适的词语填在文章中的括号里。 A．干干瘦瘦 B．稳稳当当 C．平平淡淡 D．皱皱巴巴 （2）哑巴老人的船________，他摆得________，从中可以看出哑巴老人________。 （3）“好像这一切都是理所当然的。”句中的“这一切”指的是________。 （4）哑巴老人生病住院，人们怀念他的原因有哪些？ （5）通读全文，哑巴老人一生做了哪几件事？你觉得哑巴老人是个怎样的人？ 【答案】（1）A；D；B；C （2）干净而清爽；又快又稳；他是勤快、摆渡技巧高超、做事利索 （3）年复一年哑巴老人用渡船载人过河。 （4）哑巴老人摆船又快又稳；收费便宜，渡河只收一毛钱；在这里过河，不会耽搁行程。（5）炸桥、摆渡、建桥；哑巴老人识大体、顾大局，一生默默奉献，不求回报。 【解析】【分析】（1）本题考查词语的运用。应在理解词语的基础上，结合具体语境选词填空，使句中表达更准确。 （2）本题主要考查对短文内容的理解记忆能力。解答本题，需要真正阅读短文的内容，然后结合前后句(题目提供的信息)进行补充填空即可。 （3）这是考查指代性词语指代的内容。感知文本内容，从文章中提炼和概括信息，结合 一、 二、 三、 填空题（每小题 分，共 ?分） 、设原??问题为?????? ?≥-=++-≥--≤++++-= ,0,5232 4 7 532min 3213213213213 21无约束x x x x x x x x x x x x x x x Z 则它的标准形和对偶规划问题分别为：________________________ 和 ________________________。 、用分枝定界法求整数规划12 12121121min 5 2 56 30 4,0Z x x x x x x x x x x =---≥-??+≤?? ≤??≥?且为整数 的解时，求得放松问题的解为? ＝ ? ? ? ? ? ?，则可将原问题分成如下两个子问题 与 求解。 、右图的最小支撑图是。 、右边的网络图是标号算法中的图，其中每条弧上的数 表示其容量和流量。该图中得到的可行流的增广链 (-3,1) (2,1) ②5(4) ④ ① 6(6) 6(4) ⑥ (0, ∞) 8(8) 3(2 ) 9(9)(5,1) 为： ，在其上可增的最大流量 为 。 、已知某线性规划问题，最优单纯形表如下 则其最优解为： ，最优值 max Z 。 二、单项选择题（每小题 分，共 分） 、下列表格是对偶单纯形表的是（ ? ） 、关于线性规划模型的可行域，叙述正确的为（ ） ?、可行域必有界； 、可行域必然包括原点； 、可行域必是凸的； 、可行域内必有无穷多个点。 、在运输问题中如果总需求量大于总供应量，则求解时应（ ） ?、虚设一些供应量； ?、虚设一个供应点； 、根据需求短缺量，虚设多个需求点； ?、虚设一个需求点。 、下列规划问题不可用动态规划方法求解的是（ ） ?、背包问题； ?、最短路径问题 、线性规化： ???≥≥=++++=0 ,010 34..max 321 3 32211y x x x x t s x c x c x c Z ?、22 min (,)(2)3(1).. 460,0f x y x y s t xy y x y ?=++-?+ 五年级阅读理解及答案100篇 【篇一：五年级语文阅读理解100篇】 xt>一、古诗文填空(60分) 1、《清明》的作者是，这首诗的前两句是，。 2、 3、“总把新桃换旧符。”是的诗句。 4、《示儿》的作者是，他是代诗人，诗的后两句是，。 5这样的诗句。 6、。春风又绿江南岸，。 7、。欲把西湖比西子，。 8、，花木成畦手自栽。，。 9、。卷地风来忽吹散，。 10、。二、默写古诗（20分） 1、默写两首描送别朋友的古诗： 《》作者。 。 《》作者。 。 2、默写两首带“银河”字样的古诗： 《》作者 。 。 《》作者 。 。 三、阅读积累（30分） 一、课内：（16分） 1、词，是我国古代的一种，因为句子有长有短，所 以又叫。 2、，，是知也。 3、子不学，。，。 4、“白杨树生根发芽，长出粗壮的枝干。不管遇到还是，不管遇到还是，它总是那么，那么，不，也不。” 二、课外：14分 1、写出“金陵十二钗”中的四个 人：。 2、“病补孔雀裘”的是。“痛吟《葬花词》”的是。 3、《儒林外史》的作者是 4、巴尔扎克是国著名的作家，他的作品有很多，请写 出其中的两部作品名称《》《》 5我班图书角还有他的诗集呢。 6、法国的大仲马著名的作品有《》《 【篇二：沪教版五年级升初中语文阅读理解100篇(第 二部分)】 txt>亵渎可远观而不可亵玩焉 怎样阅读写人的文章● 分析人物的描写方法 姓名班级学号计分 【知识要点】 ● 分析人物的描写方法 写人文章，是以表现人物为重点的，而人物的特点 , 除了通过事情来表现 外，还通过人物描写来展示。人物描写方法，主要有肖像描写、行动描写、语言 描写、心理描写等。只有这样 , 才能把人物写得活灵活现。我们在阅读时，一定 要抓住人物的言行举止，体会人物的个性品质理解作者所要表达的意思。 ●分析写作目的，归纳中心思想 明确了文章写什么，还要弄清作者为什么写这篇文章。对中心思想的总结 概括是加深对课文内容的理解，进一步明确人物形象的有效途径。 一、基本积累：（40分） 1、按要求写四字词语：（18分） 《管理运筹学》期中测试题 第一部分 线性规划 一、填空题 1．线性规划问题是求一个 目标函数 在一组 约束条件 下的最值问题。 2．图解法适用于含有 两个 _ 变量的线性规划问题。 3．线性规划问题的可行解是指满足 所有约束条件_ 的解。 4．在线性规划问题的基本解中，所有的非基变量等于 零 。 5．在线性规划问题中，基本可行解的非零分量所对应的列向量线性 无 关 6．若线性规划问题有最优解，则最优解一定可以在可行域的 顶点_ 达到。 7．若线性规划问题有可行解，则 一定 _ 有基本可行解。 8．如果线性规划问题存在目标函数为有限值的最优解，求解时只需在其 可行解 的集合中进行搜索即可得到最优解。 9．满足 非负 _ 条件的基本解称为基本可行解。 10．在将线性规划问题的一般形式转化为标准形式时，引入的松驰变量在目标函数中的系数为 正 。 11．将线性规划模型化成标准形式时，“≤”的约束条件要在不等式左_端加入 松弛 _ 变量。 12．线性规划模型包括 决策变量 、目标函数 、约束条件 三个要素。 13．线性规划问题可分为目标函数求 最大 _ 值和 最小 _值两类。 14．线性规划问题的标准形式中，约束条件取 等 _ 式，目标函数求 最大 _值，而所有决策变量必须 非负 。 15．线性规划问题的基本可行解与基本解的关系是 基本可行解一定是基本解，反之不然 16．在用图解法求解线性规划问题时，如果取得最值的等值线与可行域的一段边界重合，则 _ 最优解不唯一 。 17．求解线性规划问题可能的结果有 唯一最优解，无穷多最优解，无界解，无可行解 。 18.如果某个约束条件是“ ”情形，若化为标准形式，需要引入一个 剩余 _ 变量。 19.如果某个变量X j 为自由变量，则应引进两个非负变量X j ′ , X j 〞, 同时令X j ＝ X j ′ - X j 〞 j 。 20.表达线性规划的简式中目标函数为 线性函数 _ 。 21.线性规划一般表达式中，a ij 表示该元素位置在约束条件的 第i 个不等式的第j 个决策变量的系数 。 22．线性规划的代数解法主要利用了代数消去法的原理，实现_ 基变量 的转换，寻找最优解。 23．对于目标函数最大值型的线性规划问题，用单纯型法代数形式求解时，当非基变量检验数_ 非正 时，当前解为最优解。 24．在单纯形迭代中，选出基变量时应遵循_ 最小比值 法则。 二、单选题 1． 如果一个线性规划问题有n 个变量，m 个约束方程(m 五年级阅读理解题及答案 (内含45个五年级阅读专练) 深山含笑（五年级阅读训练A） 我以前见过的含笑花都是庭院种植的，叶细花小，象牙色的花蕊吐着幽香，有一种水果般的甜沁。含笑不（以、已）艳丽著称，妙的是一缕沁香。 在井冈山深处，我被另一种含笑花（佩、折）服！几株两三丈高的乔木体如游龙，散发出弥天的清香气息，这就是野生的深山含笑。多么突兀的秀色啊！她简直像一个绝世独立的北方佳人，（竟、竞）然在大山深处隐藏了如此潇洒、如此豪放的春光。和庭院含笑相比，倒（像、向）是临风挺立的 巾帼英雄，笑得那么爽朗、欢畅。那是一种胜利的喜悦，似乎天上的白云都是从她的胸中笑出来的。 从小路那边走过来两个拎着简单行李的年轻人。他们是那个边远的、还没通车的村子里的老师、跟着他们，我们也进了村。目睹孩子们围着老师的亲切嬉闹，我忽然感觉另有一株高大的深山含笑在我心中晃动起来…… 1．把文中括号里有不合适的字划掉。（5分） 2．庭院中的含笑与野生的含笑有什么不同？（3分） ________________________________________________ _________________________________________________ 3．在文中用曲线画出两个比喻句。（2分） 4．注意带点词语，结合题目写出文章最后一句话的意思。（4分） --------------- （五年级阅读训练二） 王若飞同志是一位无产阶级革命家。解放前，他因从事革命工作，被敌人逮捕了。在监狱里，他经常对难友们说：“敌人要摧残我们，我们一定要爱护自己的身体， 我们是革命者，决不能向恶劣的环境屈服，要坚决斗争。” 王若飞同志的身体不好，为了坚持对敌斗争，他想方设法，利用各种条件锻炼身体。 王若飞同志在狱中的锻炼方法之一是日光浴。他利用每天短暂的放风时间到院子里晒太阳。后来，他得了严重的风湿性关节炎，敌人被迫允许他每天晒一两小时太阳。他就利用这个机会，躺在院子里让太阳晒全身，把皮肤晒得紫红紫红的。 冷水擦身，是王若飞锻炼身体的另一种方法。那时，反动派百般折磨政治犯，别说洗澡，就连喝的水也不供给。但王若飞的言行感动了出身贫苦的老看守员，他偷偷地给王若飞买了几只大碗，王若飞同志每天用它盛冷水，用毛巾蘸着擦身，擦到全身发红为止。 王若飞同志在狱中还有一种锻炼方法，叫做“室内体操”。体操包括伸腿、弯腰、曲臂等动作。不管三九天，还是三伏天，他都坚持锻炼。 一次，一个难友问王若飞：“我有一事不明白，你骂**，骂蒋介石，天不怕，地不怕，真是好汉。可是，你坐在牢房里，还天天做操， 【阅读理解】五年级语文阅读理解精选及答案 我喜欢露珠，我赞美它。 夏日的清晨，我来到野外散步。在万道霞光中，一个美妙的，崭新的世界便（显现呈现出现）在我的眼前。看！那无边的绿叶上，闪动着无数颗晶莹的露珠，一颗、两颗、十颗、百颗……啊！像夜空中璀璨的繁星，像碧波上撒满了宝石，又像千百万闪光的眼睛。啊！露珠！这就是极普通又为人们所喜爱的露珠。 露珠的身体很小，生命也很短暂，但它却是不平凡的。当夜幕笼罩的时候它像慈母用乳汁（教育哺育培育）儿女一样地滋润着禾苗每当黎明到来的时候它又最早睁开那不知疲倦的眼睛它白天隐身于空气中，夜晚无声地辛勤工作。它不像暴雨那样挟带风雷闪电以（夸耀显耀炫耀）它的威力，更不像冰雹那样对一切（残酷残忍残暴）无情。它把短暂的一生，献给树木、庄稼……而对它们却无所苛求。它多么像辛勤的园丁，培育着祖国的花朵，多么像我们敬爱的老师，伏在桌边灯下夜以继日地工作，把全部心血洒在我们的心田…… 我爱露珠，它比珍珠更珍贵，比宝石更晶莹，比群星更璀璨。我赞美它。露珠精神将永远在我心头闪动。 １、联系上下文，在文中括号处选择恰当的词语，依次写在横线上。 ＿＿＿＿＿＿＿＿＿＿＿＿＿＿＿ ＿＿＿＿＿ ２、把文章第三自然段中没有标点的地方加上标点。 ３、联系上下文理解字义。 ⑴它白天隐身于空气中。于：＿＿＿＿＿＿＿＿＿＿＿ ⑵又为人们所喜爱。为：＿＿＿＿＿＿＿＿＿＿＿＿＿ ４、“露珠精神将永远在我心头闪动”中“露珠精神”是指＿＿＿＿＿＿＿＿＿ 2 这是一个迷人的海滨的夏夜。 夕阳西下，天空燃烧着一片橘红色的晚霞。大海也被霞光染成了红色，它比天空的景色更壮观。当一排排波浪涌起的时候，那映照在浪峰上的霞光，又红又亮，闪烁着，滚动着，使人赞叹。 夜幕降临，天空的霞光渐渐淡下去了，深红的颜色变成了浅红，当一切光都消失了的时候，那高而远的天空中便出现了夜明星。它是那么大，那么亮，放射着令人注目的光辉，活像一盏挂在高空的明灯． 夜色加浓，各处的灯火，也陆续亮了起来，尤其是山坡上那一片片灯光，它们从半空中倒映在乌蓝的海面上，像一串流动的珍珠． 我踏着软绵绵的沙滩，望着这夏夜的景色，心里有说不出的愉快和兴奋。 四 川 大 学 网 络 教 育 学 院 模 拟 试 题( A ) 《管理运筹学》 一、 单选题（每题２分，共20分。） 1．目标函数取极小（minZ ）的线性规划问题可以转化为目标函数取极大的线性规 划问题求解，原问题的目标函数值等于（ C ）。 A. maxZ B. max(-Z) C. –max(-Z) D.-maxZ 2. 下列说法中正确的是（ B ）。 Ａ．基本解一定是可行解 Ｂ．基本可行解的每个分量一定非负 Ｃ．若B 是基，则B 一定是可逆Ｄ．非基变量的系数列向量一定是线性相关的 3．在线性规划模型中，没有非负约束的变量称为 （ D ） 多余变量 B ．松弛变量 C ．人工变量 D ．自由变量 4. 当满足最优解，且检验数为零的变量的个数大于基变量的个数时，可求得（ A ）。 Ａ．多重解 Ｂ．无解 Ｃ．正则解 Ｄ．退化解 5．对偶单纯型法与标准单纯型法的主要区别是每次迭代的基变量都满足最优检验但不完全满足 （ D ）。 A ．等式约束 B ．“≤”型约束 C ．“≥”约束 D ．非负约束 6. 原问题的第ｉ个约束方程是“＝”型，则对偶问题的变量i y 是（ B ）。 Ａ．多余变量 Ｂ．自由变量 Ｃ．松弛变量 Ｄ．非负变量 7.在运输方案中出现退化现象，是指数字格的数目( C )。 A.等于m+n B.大于m+n-1 C.小于m+n-1 D.等于m+n-1 8. 树Ｔ的任意两个顶点间恰好有一条（ B ）。 Ａ．边 Ｂ．初等链 Ｃ．欧拉圈 Ｄ．回路 9．若G 中不存在流f 增流链，则f 为G 的 （ B ）。 A ．最小流 B ．最大流 C ．最小费用流 D ．无法确定 10.对偶单纯型法与标准单纯型法的主要区别是每次迭代的基变量都满足最优检验但不完全满足（ D ） Ａ．等式约束 Ｂ．“≤”型约束 Ｃ．“≥”型约束 Ｄ．非负约束 二、多项选择题（每小题4分，共20分） 1．化一般规划模型为标准型时，可能引入的变量有 （ ） A ．松弛变量 B ．剩余变量 C ．非负变量 D ．非正变量 E ．自由变量 2．图解法求解线性规划问题的主要过程有 （ ） A ．画出可行域 B ．求出顶点坐标 C ．求最优目标值 D ．选基本解 E ．选最优解 3．表上作业法中确定换出变量的过程有 （ ） A ．判断检验数是否都非负 B ．选最大检验数 C ．确定换出变量 D ．选最小检验数 E ．确定换入变量 4．求解约束条件为“≥”型的线性规划、构造基本矩阵时，可用的变量有 （ ） A ．人工变量 B ．松弛变量 C. 负变量 D ．剩余变量 E ．稳态 变量 5．线性规划问题的主要特征有 （ ） A ．目标是线性的 B ．约束是线性的 C ．求目标最大值 D ．求目标最小值 E ．非线性 三、 计算题（共60分） 1. 下列线性规划问题化为标准型。(10分) 深山含笑（五年级阅读训练一） 我以前见过的含笑花都是庭院种植的，叶细花小，象牙色的花蕊吐着幽香，有一种水果般的甜沁。含笑不（以、已）艳丽著称，妙的是一缕沁香。 在井冈山深处，我被另一种含笑花（佩、折）服！几株两三丈高的乔木体如游龙，散发出弥天的清香气息，这就是野生的深山含笑。多么突兀的秀色啊！她简直像一个绝世独立的北方佳人，（竟、竞）然在大山深处隐藏了如此潇洒、如此豪放的春光。和庭院含笑相比，倒（像、向）是临风挺立的 巾帼英雄，笑得那么爽朗、欢畅。那是一种胜利的喜悦，似乎天上的白云都是从她的胸中笑出来的。 从小路那边走过来两个拎着简单行李的年轻人。他们是那个边远的、还没通车的村子里的老师、跟着他们，我们也进了村。目睹孩子们围着老师的亲切嬉闹，我忽然感觉另有一株高大的深山含笑在我心中晃动起来…… 1．把文中括号里有不合适的字划掉。（5分） 2．庭院中的含笑与野生的含笑有什么不同？（3分） ________________________________________________ _________________________________________________ 3．在文中用曲线画出两个比喻句。（2分） 4．注意带点词语，结合题目写出文章最后一句话的意思。（4分） --------------- （五年级阅读训练二） 王若飞同志是一位无产阶级革命家。解放前，他因从事革命工作，被敌人逮捕了。在监狱里，他经常对难友们说：“敌人要摧残我们，我们一定要爱护自己的身体， 我们是革命者，决不能向恶劣的环境屈服，要坚决斗争。” 王若飞同志的身体不好，为了坚持对敌斗争，他想方设法，利用各种条件锻炼身体。 王若飞同志在狱中的锻炼方法之一是日光浴。他利用每天短暂的放风时间到院子里晒太阳。后来，他得了严重的风湿性关节炎，敌人被迫允许他每天晒一两小时太阳。他就利用这个机会，躺在院子里让太阳晒全身，把皮肤晒得紫红紫红的。 冷水擦身，是王若飞锻炼身体的另一种方法。那时，反动派百般折磨政治犯，别说洗澡，就连喝的水也不供给。但王若飞的言行感动了出身贫苦的老看守员，他偷偷地给王若飞买了几只大碗，王若飞同志每天用它盛冷水，用毛巾蘸着擦身，擦到全身发红为止。 王若飞同志在狱中还有一种锻炼方法，叫做“室内体操”。体操包括伸腿、弯腰、曲臂等动作。不管三九天，还是三伏天，他都坚持锻炼。 一次，一个难友问王若飞：“我有一事不明白，你骂**，骂蒋介石，天不怕，地不怕，真是好汉。可是，你坐在牢房里，还天天做操，又好像很爱护自己的身体，这 究竟是怎么回事？” 四川大学网络教育学院模拟试题( A ) 《管理运筹学》 一、单选题（每题２分，共20分。） 1．目标函数取极小（minZ）的线性规划问题可以转化为目标函数取极大的线性规 划问题求解，原问题的目标函数值等于（C）。 A. maxZ B. max(-Z) C. –max(-Z) D.-maxZ 2.下列说法中正确的是（B）。 Ａ．基本解一定是可行解Ｂ．基本可行解的每个分量一定非负 Ｃ．若B是基，则B一定是可逆Ｄ．非基变量的系数列向量一定是线性相关的3．在线性规划模型中，没有非负约束的变量称为（ D ） 多余变量B．松弛变量C．人工变量D．自由变量 4. 当满足最优解，且检验数为零的变量的个数大于基变量的个数时，可求得 （ A ）。 Ａ．多重解Ｂ．无解Ｃ．正则解Ｄ．退化解5．对偶单纯型法与标准单纯型法的主要区别是每次迭代的基变量都满足最优检验 但不完全满足（ D ）。 A．等式约束 B．“≤”型约束 C．“≥”约束 D．非负约束 y是（ B ）。 6. 原问题的第ｉ个约束方程是“＝”型，则对偶问题的变量i Ａ．多余变量Ｂ．自由变量Ｃ．松弛变量Ｄ．非负变量 7.在运输方案中出现退化现象，是指数字格的数目( C )。 A.等于m+n B.大于m+n-1 C.小于m+n-1 D.等于m+n-1 8.树Ｔ的任意两个顶点间恰好有一条（B）。 Ａ．边Ｂ．初等链Ｃ．欧拉圈Ｄ．回路9．若G中不存在流f增流链，则f为G的（ B ）。 A．最小流 B．最大流 C．最小费用流 D．无法确定 10.对偶单纯型法与标准单纯型法的主要区别是每次迭代的基变量都满足最优检验 但不完全满足（ D ） Ａ．等式约束Ｂ．“≤”型约束Ｃ．“≥”型约束Ｄ．非负约束二、多项选择题（每小题4分，共20分） 1．化一般规划模型为标准型时，可能引入的变量有（） A．松弛变量 B．剩余变量 C．非负变量 D．非正变量 E．自由变量 2．图解法求解线性规划问题的主要过程有（） A．画出可行域 B．求出顶点坐标 C．求最优目标值 D．选基本解 E．选最优解 3．表上作业法中确定换出变量的过程有（） A．判断检验数是否都非负 B．选最大检验数 C．确定换出变量 D．选最小检验数 E．确定换入变量 4．求解约束条件为“≥”型的线性规划、构造基本矩阵时，可用的变量有（）A．人工变量 B．松弛变量 C. 负变量 D．剩余变量 E．稳态变量 5．线性规划问题的主要特征有（） A．目标是线性的 B．约束是线性的 C．求目标最大值 D．求目标最小值 E．非线性 三、计算题（共60分） 1. 下列线性规划问题化为标准型。(10分) 1 / 17 502班暑假学俱乐部阅读内容 1. 风雪夜中的一盏灯 我家对面有一座山，山腰上有一所学校，童年，我就在这里上学。 隔着弯弯曲曲的小河和绿油油的稻田，我的窗户正好对着学校里张老师的窗户，我常常喜欢向那里眺望。每晚，那个窗口都闪烁着灯光。 那一年的冬天来得特别早天气也格外冷一个风雪交加的夜晚我早早 就钻进了铺得软软的被窝一觉醒来我又习惯地向对面山腰望去透过纷纷扬扬的雪花张老师的窗口像往常一样闪烁着灯光在这沉沉的雪夜这灯光显得分外明亮、耀眼，我猛然想起，张老师的木柴已经烧光了！这样冷的天，张老师拿什么取暖？我急忙翻身下床，悄悄爬上小楼阁，把大哥留着大年三十炖猪头的好木柴"背"了一捆，向对面山腰跑去。 赶到老师窗下，我被眼前的情景惊呆了：寒风夹着雪花顺着窗棂的缝隙不住往里灌，年久失修的窗扇也在风中不住颤抖着。张老师握着红笔，正在批改作业。他不停地跺脚，还不时放下笔来，往手里哈着热气，灯光照着他清瘦的面庞和冻得发青的嘴唇，他的鼻尖和耳朵冻得通红…… 看着这一切，我的心里就像吞进了铅块，难受极了。我贴着窗户，轻轻地叫了一声："张老师"。他吃惊地抬起头来，认出是我，便急忙打开了门……回到家，我又钻进被窝，还做了一个梦：我梦见张老师穿着厚厚的新棉衣，做在红红的炭火旁给我们批改作业，他的眼睛闪着光，脸上露出了笑容…… 学校门前的桃花开了又落，落了又开，一批批同学来了又去，去了又来。张老师窗口的灯光却无论风霜雨雪，暑往寒来，每天都亮到深夜。他是在用青春和心血点燃着知识的明灯，照亮我们前进的路…… 1、写出下列词语的反义词。2分 喜欢（）前进（）寒冷（）急忙（）2、写出下列词语的近义词。2分 分外（）吃惊（）明亮（）往常（） 3、给文章第3自然段前部分加上标点。3分 4、选择正确的答案，（在序号上打"√"）2分 （1）一个风雪交加的夜晚，"我"为张老师送柴，说明"我"尊敬关心老师。 （2）一个风雪交加的夜晚，张老师伏案批改作业，说明张老师忠诚于党的教育事业。 5、摘录文中具体描写张老师冬夜冒着严寒为学生批改作业的一句话。圈出文中反映张老师冷得厉害的词语。2分 _______________________________________________________ ___ _____________________________________________________ _____ 6、理解句子的含义，将正确答案的序号填在括号里。4分 （1）"看着这一切，我的心里就像吞进了铅块，难受极了。"我难受是因为（） a、张老师冷得厉害。 b、张老师工作到深夜，实在太辛苦了。 五年级语文阅读理解答题技巧 一、做阅读理解的三个步骤。 1、通读全文，掌握大意。在解答阅读理解时，同学们先要快速地浏览一下整篇文章，重视标题、开头段、结尾段及各段落的首句，理清脉络，了解基本梗概，不要把时间花在生词难句上。每认真读完一段，要及时概括段意。 2、认真审题，定向阅读。在掌握文章的大意之后，认真读短文后面的题目的每个字，然后带着这些问题仔细的阅读第二遍，以做到有目的地阅读，并在原文找出大体范围，就等于答对该题的一半了。 3、复读全文，验证答案。答题完毕后，同学们应对照答案将整篇文章从头到尾再看一遍，以确保答案的正确。同时答案要做到：准确、简洁、全面。 二、阅读理解答题技巧点拨。 1、概括段落大意 （1）要准确地概括出段意，首先要读懂段落每句话的意思，还要弄清楚段内各句的相互关系，找出能揭示全段意思的主要句子，即所谓的中心句（中心句的位置多数在段首或段末，个别也有在段中的）。如果没有中心句的，就要抓住全段的中心意思，自己总结概括。 （2）摘句法，即找出段落中的中心句。 （3）概述法，用自己的语言概述全段的主要意思。 （4）联合法。有些段落讲的不止一个意思，概括时必须用简练的语言把几个意思表述出来，缺一不可，这就是联合内容要点加以概述。 2、怎样概括文章的中心思想 文章的中心思想就是作者的写作意图、目的。它是通过文章的字、词、段、篇的结构形式表达出来的。我们要归纳中心思想，首先必须读懂文章的主要内容、段意或文章的中心句来概括总结，也可以从审题、文章的开头、结尾、重点段、议论部分或从考题中得到提示入手。其主要方法有： （1）先概括文章的主要内容，再想一想作者为什么要写这些内容，然后领会写作目的，即中心思想。如《穷人》主要描写了渔夫和他的妻子桑娜，不管自己家境困难依然收养了自己已故邻居的两个孤儿这件事，想一想作者为什么要写这些呢，从文中可以分析出作者的写作目的是：反映沙俄时代渔民的悲惨生活和穷人宁可自己吃苦，也要互相帮主的高尚品质。 （2）分析课文的重点段。从文中找出中心思想，如《卖火柴的小女孩》的第二段，描写了卖火柴的小女孩几次擦燃火柴所产生的美好的幻觉以及幻觉瞬间消逝，重新回到冷酷的现实，再联系第一段和第三段“惨死接头”就不拿找出中心：作者的目的就是要揭露资本主义社会的罪恶，表达了他对小女孩的不幸遭遇的深切同情。 （3）从文题找中心，有些文章的题目就直接点明了中心。如《伟大的友谊》，文章歌颂了马克思和恩格斯的伟大的革命友谊。 （4）从中心句找中心，如《我的伯父鲁迅先生》中最后一句：“伯父就是这样一个人，他为别人想得多，为自己想的少。”又如《我的心事》中反复出现的一句话：“说话要算话”就是这篇文章的中心。 3、景物描写的作用——交代故事发生的时间、地点；渲染气氛，烘托人物心情；表现人物性格；推动情节的发展。 4、“根据语境解释词语”题型 有两种答题方式，一种是通过对上下文的分析（即联系上下文），直接写出该词语的意思；另一种更为完整，即在解释完该词语后，再加上——在文中指的是…… 5、“赏析优美语句”题型（主要有两种题型） Ａ、“××”词好在哪里？ 答题方式：用了“××”词，生动地（准确地）说明了……事物的……特征，能够激发读者的兴趣（符合实际情况，具有科学性）。 Ｂ、“××”词能不能删掉？ 答题方式：①不能，用了“××”词，生动地说明了……，能够激发读者的 《运筹学》试题样卷（一） 一、判断题（共计10分，每小题1分，对的打√，错的打X ） 1. 无孤立点的图一定是连通图。 2. 对于线性规划的原问题和其对偶问题，若其中一个有最优解， 另一个也一定有最优解。 3. 如果一个线性规划问题有可行解，那么它必有最优解。 4．对偶问题的对偶问题一定是原问题。 5．用单纯形法求解标准形式（求最小值）的线性规划问题时，与0 >j σ对应的变量都可以被选作换 入变量。 6．若线性规划的原问题有无穷多个最优解时，其对偶问题也有无穷 多个最优解。 7. 度为0的点称为悬挂点。 8. 表上作业法实质上就是求解运输问题的单纯形法。 9. 一个图G 是树的充分必要条件是边数最少的无孤立点的图。 二、建立下面问题的线性规划模型（8分） 某农场有100公顷土地及15000元资金可用于发展生产。农场劳动力情况为秋冬季3500人日；春夏季4000人日。如劳动力本身用不了时可外出打工，春秋季收入为25元 / 人日，秋冬季收入为20元 / 人日。该农场种植三种作物：大豆、玉米、小麦，并饲养奶牛和鸡。种作物时不需要专门投资，而饲养每头奶牛需投资800元，每只鸡投资3元。养奶牛时每头需拨出1.5公顷土地种饲料，并占用人工秋冬季为100人日，春夏季为50人日，年净收入900 元 / 每头奶牛。养鸡时不占用土地，需人工为每只鸡秋冬季0.6人日，春夏季为0.3人日，年净收入2元 / 每只鸡。农场现有鸡舍允许最多养1500 三、已知下表为求解某目标函数为极大化线性规划问题的最终单纯形表，表中54 ,x x 为松弛变量，问 (1)写出原线性规划问题；（4分） (2)写出原问题的对偶问题；（3分） (3)直接由上表写出对偶问题的最优解。（1分） 四、用单纯形法解下列线性规划问题（16分） s. t. 3 x1 + x2 + x3?60 x 1- x 2 +2 x 3?10 x 1+x 2-x 3?20 x 1,x 2 ,x 3?0 五、求解下面运输问题。（18分） 某公司从三个产地A1、A2、A3将物品运往四个销地B1、B2、B3、B4，各产地的产量、各销地的销量和各产地运往各销地每件物品的运费如表所示： 问：应如何调运，可使得总运输费最小? 六、灵敏度分析（共8分） 线性规划max z = 10x1 + 6x2 + 4x3 s.t. x1 + x2 + x3 ?100 10x1 +4 x2 + 5 x3 ?600 2x1 +2 x2 + 6 x3 ?300 x1 , x2 , x3 ?0 的最优单纯形表如下： (1)C1在何范围内变化，最优计划不变？(4分) (2)b1在什么范围内变化，最优基不变？(4分) 七、试建立一个动态规划模型。（共8分） 鸵鸟分布于非洲和阿拉伯半岛，主要生活在沙漠地区，是现代鸟类中体形最大的鸟。 鸵鸟的双翅已经退化，不会飞翔，但在顺风及拐弯奔跑时，双翅展开，起着像帆一样的作用，能保持平衡，又能借助强劲风力，加快奔跑速度。鸵鸟善于奔跑，一步可以跨出7米远，时速可达60千米。当然它们不能一天到晚都跑得这么快，但即使在它们跑得疲倦时，也能胜过一匹快马。 提到鸵鸟，人们往往认为，鸵鸟遇到危险的时候，如果来不及逃跑，就会把自己的头和颈平贴地面，钻进沙堆，以为自己什么也看不见，就会平安无事。其实，这是人们误解了鸵鸟。经过研究人员多次观察，从来没见过鸵鸟把头藏进沙里。实际上，鸵鸟不会这样愚蠢，因为把头埋进沙里，很快就会窒息死亡。有时鸵鸟把头贴近地面，是为了听远处的声音，或者是为了放松一下颈部的肌肉。年幼体弱的小鸵鸟，遇到敌害逼近时，常把身体紧贴地面。由于它们身上羽毛的颜色和黄草、黄沙相似，很容易瞒过敌害，即使在这种情况下，小鸵鸟的头颈也不钻进沙堆，而是两眼紧紧盯住敌害，做好随时逃离的准备。 1．文章介绍了鸵鸟的哪些特点？ ______________________________________________________________ _____________ 2．鸵鸟真的会把头藏进沙里吗？为什么？ 3．作者用了哪些说明方法描写鸵鸟善于奔跑？举例说一说。 方法一：____________例句： __________________________________________________ 方法二：____________例句： __________________________________________________ 方法三：____________例句： __________________________________________________ 运筹学期末试卷（B卷） 系别：工商管理学院专业：考试日期：年月日姓名：学号：成绩： 1．[10分] 匹克公司要安排4个工人去做4项不同的工作，每个工人完成各项工作所消耗的时间（单位：分钟）如下表所示： 要求：（1）建立线性规划模型(只建模型,不求解) （2）写出基于Lindo软件的源程序。 2.[15分]某公司下属甲、乙两个厂，有A原料360斤，B原料640斤。甲厂用A、B两种原料生产x1,x2两种产品，乙厂也用A、B两种原料生产x3,x4两种产品。每种单位产品所消耗各种原料的数量及产值、分配等如下 （1） 建立规划模型获取各厂最优生产计划。 （2） 试用图解法 求解最优结果。 3．[10分] 考虑下面的线性规划问题: 目标函数：Min Z=16x 1+16x 2 +17x 3 约束条件： 利用教材附带软件求解如下: **********************最优解如下************************* 目标函数最优值为 : 148.916 变量 最优解 相差值 ------- -------- -------- x1 7.297 0 x2 0 .703 x3 1.892 0 约束 松弛/剩余变量 对偶价格 ------- ------------- -------- 13123123123300.56153420,,0 x x x x x x x x x x x +≤-+≥+-≥≥ 1 20.811 0 2 0 -3.622 3 0 -4.73 目标函数系数范围: 变量下限当前值上限 ------- -------- -------- -------- x1 1.417 16 16.565 x2 15.297 16 无上限 x3 14.4 17 192 常数项数范围: 约束下限当前值上限 ------- -------- -------- -------- 1 9.189 30 无上限 2 3.33 3 15 111.25 3 -2.5 20 90 试回答下列问题： （1）第二个约束方程的对偶价格是一个负数(为-3.622)，它的含义是什么? （2）x2有相差值为0.703,它的含义是什么? （3）请对右端常数项范围的上、下限给予具体解释，应如何应用这些数 《管理运筹学》期末考试试题 一、单项选择题（共5小题，每小题3分，共15分） 1.如果一个线性规划问题有n个变量，m个约束方程(m 3. 写出下面线性规划问题的对偶问题： 123123123123123min z 25, 258, 23 3,.. 4 26, ,,0. x x x x x x x x x s t x x x x x x =++-+≤??++=??-+≤??≥? 四、计算下列各题（每题20分，合计40分） 1. 用单纯形法求解下列线性规划的最优解： 012121212max 2..32250,0x x x s t x x x x x x =+??≤??≤??+≤??≥≥? 2.用割平面法求解整数规划问题。 12 121212 max 7936735,0,z x x x x x x x x =+-+≤??+≤??≥?且为整数 心田上的百合花 林清玄 在一个偏僻遥远的山谷里，有一个高达数千尺的断崖。不知道什么时候，断崖边上长出了一株小小的百合。百合刚刚诞生的时候，长得和杂草一模一样。但是，它心里知道自己并不是一株野草。它的内心深处，有一个内在的纯洁的念头：“我是一株百合，不是一株野草。惟一能证明我是百合的办法，就是开出美丽的花朵。” 有了这个念头，百合努力地吸收水分和阳光，深深地扎根，直直地挺着胸膛。终于在一个春天的清晨，百合的顶部结出了第一个花苞。 百合的心里很高兴，附近的杂草却很不屑，它们在私底下嘲笑着百合：“这家伙明明是一株草，偏偏说自己是一株花，还真以为自己是一株花，我看它顶上结的不是花苞，而是头脑长瘤了。” 开场合，它们则讥讽百合：“你不要做梦了，即使你真的会开花，在这荒郊野外，你的价值还不是跟我们一样？” 百合说：“我要开花，是因为我知道自己有美丽的花；我要开花，是为了完成作为一株花的庄严使命；我要开花，是由于喜欢以花来证明自己的存在。不管有没有人欣赏，不管你们怎么看我，我都要开花！” 在野草的鄙夷下，野百合努力地释放内心的能量。有一天；它终于开花了，它那灵性的白和秀挺的风姿，成为断崖上最美丽的风景。这时候，野草与蜂蝶再也不敢嘲笑它了。 百合花一朵一朵地盛开着，花朵上每天都会出现晶莹的水珠，野草们以为那是昨夜的露水，只有百合自己知道，那是极深沉的欢喜所结的泪滴。年年春天，野百合努力地开花、结籽。它的种子随着风，落在山谷、草原和悬崖边上，到处都开满洁白的野百合。 几十年后，远在百里外的人，从城市，从乡村，千里迢迢赶来欣赏百合开花。许多孩童跪下来，闻嗅百合花的芬芳；许多情侣互相拥抱，许下了“百年好合”的誓言；无数的人看到这从未见过的美，感动得落泪，触动内心那纯净温柔的一角。 不管别人怎么欣赏，满山的百合花都谨记着第一株百合的教导： “我们要全心全意默默地开花，以花来证明自己的存在。” 阅读训练： 一、写出同一偏旁的词语。（写出五个即可） 例：偏僻、遥远 五年级阅读专项练习题 姓名 （一）＿＿____＿____＿_＿_____ 新春佳节之际，妈妈买来了一幅画。我粗粗一看,这幅画并没有什么特别吸引人的地方。我心想：书店里有那么多色彩鲜艳的年画,妈妈为什么要买这幅呢?我一边琢磨,一边细细欣赏起来。 这是一幅国画，画的背景是巍峨挺拔的高山,被淡淡的烟云笼罩着,使群山若隐若现,令人格外神往。山坡下横插过来一株苍劲的松树,松树下是一对飞奔的骏马。奔在前头的是一匹枣红马,它浑身火红,仰天长嘶，红色的鬃毛高高飘起，全身的肌肉结实得像拳击场上的运动员,一块一块地凸出来，显得十分健美有力。枣红马腾空而起,柔软而漂亮的马尾巴甩得很起劲。紧挨着枣红马旁边是一匹白马,从头到尾一片洁白。连一根杂毛也没有，像一团白云轻轻飘来，显得那么洒脱矫健,分明是一匹千里马。 我越看越着迷,似乎听到了它们“得得”的马蹄声,看到了他们身后卷起的阵阵尘土,仿佛看见它们自由自在地飞驰在广阔的原野上。 这时,我猛然想起刚才的疑问，便跑去问妈妈。妈妈神秘地笑了，反问我：“你属什么呀？”我恍然大悟,抢着说：“知道了，知道了，我属马,这年画上的马象征着我。您希望我努力学习，不怕挫折，永远向前飞奔，对吗?”妈妈满意地点点头。 啊!马儿,你知道这是妈妈对我殷切的期望，我决心要让它变成现实。 1．填空。 (１)这幅国画画的是( ) 、（)、（)、（ )，主要画的是（ )。 (2)妈妈买这幅画的目的是( ）。 ２.用横线画出写“我”观画时联想的句子。 3．用‖把文章分成３段,用简洁的语句写出文章的叙述顺序。 ______ ____＿＿__＿_______＿____＿_＿_＿_______＿_ 4．写出文章的中心思想。 ___＿__＿_＿_＿_____＿＿__＿_＿_________＿_＿_____ 5．给短文加标题。 （二)深山含笑 我以前见过的含笑花都是庭院种植的，叶细花小,象牙色的花蕊吐着幽香,有一种水果般的甜沁。含笑不(以、已)艳丽著称，妙的是一缕沁香。 在井冈山深处，我被另一种含笑花（佩、折)服！几株两三丈高的乔木体如游龙,散发出弥天的清香气息,这就是野生的深山含笑。 多么突兀的秀色啊！她简直像一个绝世独立的北方佳人，（竟、竞)然在大山深处隐藏了如此潇洒、如此豪放的春光。和庭院含笑相比,倒(像、向）是临风挺立的巾帼英雄，笑得那么爽朗、欢畅。那是一种胜利的喜悦,似乎天上的白云都是从她的胸中笑出来的。 从小路那边走过来两个拎着简单行李的年轻人。他们是那个边远的、还没通车的村子里的老师、跟着他们,我们也进了村。目睹孩子们围着老师的亲切嬉闹，我忽然感觉另有一株高大的深山含笑在我心中晃动起来……【部编版五年级下册语文】阅读理解试题(含答案)
管理运筹学期末试卷B
五年级阅读理解及答案100篇-五年级阅读理解100篇
管理运筹学期中复习题答案
五年级阅读理解题及答案
【阅读理解】五年级语文阅读理解精选及答案
管理运筹学模拟试题及答案
(完整)五年级阅读训练题10篇(附答案)
管理运筹学模拟试题附答案
(完整)人教版五年级语文课外阅读及答案
五年级语文阅读理解答题技巧 含答案
《运筹学》期末考试试卷A答案
五年级阅读理解题目及答案
管理运筹学期末试卷题目B卷
《管理运筹学》期末考试试题
五年级语文阅读理解及答案
小学五年级语文阅读理解及答案
相关主题
文本预览