Approximate Grammar for Information Extraction
Abstract
In this paper, we present the concept of Approximate grammar and how it can be used to extract information from a document. As the structure of informational strings cannot be defined well in a document, we cannot use the conventional grammar rules to represent the information. Hence, the need arises to design an approximate grammar than can be used effectively to accomplish the task of Information extraction. Approximate grammars are a novel step in this direction. The rules of an approximate grammar can be given by a user or the machine can learn the rules from an annotated document. We have performed our experiments in both the above areas and the results have been impressive.
Introduction
This paper proposes the idea of approximate grammars and its usage towards information extraction. An informational string is a sequence of words that convey definite information. For example: The string “India won by 4 wickets at Old Trafford” tells us about a match result. A document contains several informational strings that should be extracted. For example: a document on cricket would contain several informational strings like “the match would be held tomorrow at Lords”, “the India-Pakistan match ended in a draw”, “Sachin Tendulkar made 67 in 60 balls”, etc.
The information extraction system consists of the following two stages,
1. Extraction of atomic information or probe. This atomic information consists of annotated texts, which fall into categories specific to the domain in consideration. For example: player-name, runs, wickets, venue, etc. could be the probes for Cricket.
2. Getting informational strings from the atomic
information or grouping. A document from which information is to be extracted is given to the probe, it first tags atomic information. (Sangal and Bansal, 2001)
(Fig1. Information extraction system)
The second stage groups the atomic information into several groups, which represent the extracted information. To group these atomic units, we are required to know the structure of the document and have the knowledge of all the possible informational strings. The permitted informational strings can be captured by a grammar.
When a document is parsed according to the rules of the grammar, several parses are possible. A normal context-free grammar does not provide a mechanism to select the most appropriate parse from all the
possible parses. To make that possible, we need to attach a probability with every rule of our grammar. Also, in an information string we can have redundant words, words that do not contribute to the informational string. These words can be called noise. In an informational string, apart from the normal atomic information or probes, we can also have patterns that can help us to extract the informational string. These patterns can be called atomic patterns.
In a document the informational units are usually close to each other. The maximum length of noise gives us the maximum possible distance between two informational units.
From all the above specifications that are useful to extract information, arises the need to have a new type of grammar. Approximate grammar incorporates all the above features and gives us the power, which can be used to do information extraction efficiently.
Structure of Approximate Grammar
At this point, one could visualize approximate grammar as a set of probabilistic CFG rules with additional features, which are explained in the later part of the paper. The rule of an approximate grammar can take the following form.
non-terminal = terminals | non-terminals | noise | probability of the rule
terminal: probe | string | root
“|” denotes OR.
Features of Approximate Grammar
Probability of a Rule:
Probability associated with a particular rule helps the information extraction system to resolve the ambiguity that arises because of several possible parses. For example, let’s take a part of a document,
“
Say, we have two rules:
1> IMP -> {name} made {runs} {location}
2> IMP -> {location} saw {runs} {wickets}
IMP: Important informational string
The above rules would give us two parses.
The probabilities of the rules attached to each of the above rule would give us the score of the final parse. The parse with the highest score would be selected to give the appropriate informational strings.
Atomic strings or Words:
The atomic strings or words make a rule more specific and thus increases the precision with which an informational string is extracted.
For example, considering a part of the document: “
The above string would be extracted using both the following rules,
IMP -> {name} {balls}
IMP -> {name} bowled {balls}
But the second rule is stricter than the first rule and represents the information more precisely.
Noise:
Noise is defined as the pattern in an informational string, which has no relevance other than fact that it appears with the atomic units. So, between any two atomic units of an informational string, noise is permitted by the grammar.
For example:
Let, SN -> noise
Then, we give a rule as
IMP -> {name} SN bowled SN {balls}
We can specify the minimum and maximum lengths
of the noise; this tells us how near or far can the two atomic units or atomic patterns in an informational string are. Roughly speaking, noise cushions the occurrence of the terminals and non-terminals of a grammar production.
So, if
SN -> noise; minlength=0, maxlength=20
Then, there could be a maximum of 20 characters between {name} and “bowled” as well as between “bowled” and {balls} in the above rule as shown seperated by a semicolon.
Terminals:
The terminals can be divided into either of the following categories. The attribute ‘cat’ denotes the category.
a) Probe: Type of the annotated strings
representing the atomic information.
Example:
b) String: A particular string the user is
expecting to be in the extracted information.
Example:
c) Root: A special type of type “string” where
the user can specify the root-word. Words in
the target document having the
morphological root as mentioned fall into this
category.
Example: ”hit” | ”score” > Then the rule: IMP-> player_name score cent; where ‘player_name’, ‘score’ and ‘cent’ are the category names of the terminals as shown above, would capture the informational string, “ Note: For the sake of simplicity, NOISE is not mentioned in the above rule. When NOISE is not mentioned, it is assumed to be there by the parser. Scoring a Parse: Each terminal is associated with a particular confidence with which it has been extracted. Scoring a parse tree takes into account the probability of the rule and the scores of its child nodes. Score of a terminal = Confidence with which a terminal is extracted. Score of a node of a parse tree = Prob. Of rule * Average score of Child nodes Example: Given P-CFG’s S-> NP VP ; prob=0.9 NP -> det n ; prob=0.7 VP -> v det n ; prob=0.6 And for the sentence “The old man the boats” Rules from the dictionary Det -> “the” ; conf= 0.9 Adj -> “old” ; conf=0.6 N -> “old” ; conf=0.3 N -> “man” ; conf=0.7 V -> “man” ; conf=0.25 N -> “boats”; conf=0.8 Generated Tree (Figure2. Generated Parse Tree for the given example) Advantages of using Approximate Grammars Resolving Ambiguity: The score of a given parse gives the information extraction system, the power to select the appropriate parse from the several possible parses. The following examples explain the above: ? Consider a piece of document, “B.A. Ambedkar is highly honored in India.” The probes would tag the above in two possible ways. One interpretation is, “ highly honored in ” Note that the production, IMP -> {degree} honored {country} ; is not a faulty production because it is valid for the following sentence: “ is honored in . Another interpretation is, “ honored in After training a sufficiently large corpus, one would find out that the probability of the production: IMP->{name}honored {country} would have a higher probability than the production: IMP- >{degree}honored{country} Then, the first interpretation of the probes is taken to be more appropriate, assuming that the confidence of extraction of atomic units is the same. ? Given the rules, IMP -> {country} {name} Prob = 0.1 IMP -> {name} {country} Prob = 0.3 Then for the piece of document, “ for sponsored by The informational string would be Sachin Tendulkar plays for India. Proximity of Atomic Units: In a document, the smaller units that build a bigger unit should be close to each other. The measure of maximum length of noise helps the system to find for units that are close to one another. For example: Given the rules, SN -> noise; minlength=0, maxlength=10 IMP -> {name} SN {runs} SN {location} And the piece of a document, “ The above rule will not extract any string from the above piece of document because the length of noise between the runs and location is 22 while the maximum permitted distance is 10. Machine Learning: The machine can be used to learn the rules of approximate grammar by analyzing an annotated text. The probability of the rules and the maximum possible noise between any two atomic units is also learned. These learned rules are then applied to an unannotated text and the informational strings corresponding to the learnt rules are extracted. Implementation of Approximate Grammars Chart Parsing: Chart Parsing is employed to parse the text according to the rules mentioned. A bottom-up approach is implemented. The algorithm of chart parsing provides an efficient way of obtaining multiple parses The advantages of using Chart Parsing are: ? It avoids multiplication of effort. ? It provides a compact representation for ‘local ambiguity’ ? It provides a representation for ‘partial parses’ Moreover, chart parsing provides a general framework in which alternative parsing and search strategies can be compared. Machine Learning to generate approximate grammars In this part of the paper, we discuss how we achieved the generation of an approximate grammar by learning patterns from a set of tagged documents. Depending on the value of a parameter, the generated rules can be mapped to the extent of generality of the patterns exhibited in the document. a) Clustering of the patterns As a primary step, the patterns of the informational strings given are clustered on the same sequence of their tags. Now, for each individual cluster, we aim to generate probabilistic grammar rules whose sum adds up to 1. b) Generating grammars for a specific cluster Now that we have the set of informational strings following a specific pattern of occurrence of the atomic probes, we compute the TRI-data for the cluster. This data contains the TRI-frequency of every consecutive atomic probes. The TRI-frequency can be defined as the number of times the sub-pattern by taking combinations of the various tag offsets found in the cluster. Note that , in the above mentioned general pattern, “string” can be null. Consider the following example, 1. 2. 3. Generated CFG’s 1.IMP -> 2.IMP -> 3.IMP -> 4.IMP -> 5.IMP -> 6.IMP -> 7.IMP -> Note: For the sake of simplicity, NOISE is not mentioned in the above rule. When NOISE is not mentioned, it is assumed to be there by the parser. (Fig 3. Generating P-CFG’s by machine learning) RHO: An important parameter in P-CFG generation, which defines a threshold below which the observed patterns in question will be discarded because of their low frequency. Disposal of such patterns is important because these patterns give rise to ambiguous productions, thereby increasing the execution time. We can extract the CFG’s corresponding to the patterns observed based on a scale of generality. This can be done by filtering the TRI’s above a threshold “RHO”. Increasing the value of this parameter results in the generation of the patterns, which are more prevalent in the texts observed. Based on the tagged documents fed as input and the nature of the extracted information, we can fine-tune this parameter to achieve more concise and relevant results. Manually specifying the rules The user has the flexibility to express his idea of the information, which is relevant to him through this framework of approximate grammars. In this way, the user can instruct the machine to specifically extract a particular pattern with ease. Experiments and Results Experiments were carried out in the domain of Cricket. The texts were taken from https://www.doczj.com/doc/0d5670728.html, website. For obtaining the annotated texts, we have made use of filters which were developed on the cricket domain. The different filters used are : player_name, runs , wickets, balls , place, tournament, team. Training was done on 80%, and testing on the remaining 20%. We were particularly interested in the outcome and other statistical details of the matches (individual scores of individual players, bowling performances of bowlers, partnerships, results etc.) and so we considered only the annotated parts where such information was present. Comments and other trivial informational strings regarding players were considered unimportant in our experimentation. Evaluation Machine generated rules captured the essential information in the domain of cricket. Table 1. illustrates the rules specifying just the probes as non-terminals on the right-hand side of the production. Stricter versions of some of the rules might be having strings/roots as the terminals on the right hand side of the production and are listed at the end of the table 1. The probabilities of the rules mentioned are not stated because they assume different values depending on the value of RHO. Note: For the sake of simplicity, NOISE is not mentioned between the probe non-terminals of the rules given below, but it is assumed to be there by the parser. IMP: Extracted rules. IMP-> team team runs runs tment venue IMP-> team player wickets runs player team runs IMP-> player runs balls balls team player IMP-> runs balls player wickets player player wickets IMP-> team runs wickets player runs player runs IMP-> team runs team venue IMP-> team venue team runs player wickets runs player wickets runs IMP-> team runs player runs player runs IMP-> player runs runs team runs IMP-> team team wickets runs player runs runs IMP-> player wickets runs player wickets runs team runs IMP-> team runs wickets team player runs IMP-> player runs player runs team balls IMP-> player runs runs player runs team runs wickets team tment venue IMP-> player runs team wickets player player IMP-> player runs player runs team runs runs runs IMP-> player team tment runs balls team runs team runs IMP-> player team runs wickets player player wickets IMP-> player balls runs IMP-> player runs player runs IMP-> player wickets IMP-> player “haul” team “beat” team IMP-> player “help” team “beat” team IMP-> team “regroup” player “bowl” player “ball” IMP-> team “regroup” player “paceman” player “ball” IMP-> team “open” player “bowl” player “ball” IMP-> player “hit” runs “keeper” player “wack” runs (Table 1: Machine generated rules for the domain of cricket) Evaluation Metric: Correctly extracted INF Precision = ------------------------------ Total extracted INF Correctly extracted INF Recall = -------------------------------------- Total number of correct INF INF: Informational Strings RHO Precision Recall 0.30 78 98.6 0.25 77.8 100 0.20 76.4 99.1 0.15 75.4 99.3 (Table 2: Results of the experiments) Table.2 shows the results obtained after extracting the informational strings in the rest 20% of the corpus. Error Analysis The misclassification of the probes in the training set brought down the precision to some extent. Consider the piece of misclassified text: The above annotated text would generate the production: IMP->player_name place player_name; And eventually, the above production would extract the informational string: The extracted informational string was not considered to be a valid one because we were specifically looking for outcome and other statistical details in a match played. By increasing the value of RHO, we could eliminate the faulty productions generated after machine learning of the training set. Unfortunately, in this process we also lost some important productions because the frequency of some patterns in the training set which occurred in the testing set were very less.Cases where RHO value was less, some valid informational strings were later discarded in the parsing process because of their low score value in the parsing process. This is because another ambiguous competitor of that production had higher frequency in the training set. Conclusion In this paper, we have proposed a framework to extract information efficiently and conveniently, considering the user’s specifications and the loosely structured document in view. The machine learning algorithm to learn the rules of approximate grammars is giving good results as experimented. Future Work We would like to categorize the informational strings into separate categories and train the system to obtain the Approximate Grammar rules for each of the categories. This would enable us to identify the category of the resultant Parse and thus would be more precise. Acknowledgement Details to be given after blind reviewing is over. References ? Briscoe, T. & Carroll, J. [1996], ``Automatic Extraction of Subcategorization from Corpora,''MS , https://www.doczj.com/doc/0d5670728.html,/users/ejb/. ? Carroll, Glenn and Eugene Charniak, Split- Corpus Grammar Learning from Tagged Texts, Computer Science Department Tech Report CS-94-08. ? Carroll, Glenn and Eugene Charniak, Learning Probabilistic Dependency Grammars from Labeled Text, Proceedings from AAAI Fall Symposium on Probabilistic Approaches to Natural Language, AAAI Press, Cambridge, MA 1992. ? Carroll, Glenn and Eugene Charniak, Two Experiments on Learning Probabilistic Dependency Grammars from Corpora, Proceedings from AAAI-92 Workshop Program, AAAI Press, San Jose, CA, 1992. ? Charniak, E.[1993], Statistical Language Learning,MIT, Cambridge, MA."Parsing with Context-free Grammars and Word Statistics", Department of Computer Science,Brown University, Technical Report CS-95-28. ? S.Soderland “CRYSTAL: Learning Domain- Specific Text Analysis Rules”. ? Dan Klein and Christopher D. Manning. An O(n3) Agenda-Based Chart Parser for Arbitrary Probabilistic Context-Free Grammars. Stanford Technical Report dbpubs/2001-16, 2001. ? Daniel H. Younger. 1967. Recognition and parsing of context free languages in time n3 . Information and Control, 10:189--208. ? Doug Arnold- “Chart Parsing” ? W.G.Lehnert, C.Cardie, D.Fisher, J.McCarthy, E.Riloff, S.Soderland - “Evaluating an Information Extraction System”. ? Fabio Ciravegna ,“Learning to tag for Information Extraction from Text” . ? S.Soderland- “Learning Text Analysis Rules for Domain Specific Natural Language Processing” . ? Roman Yangarber and Ralph Grishman. -“Machine Learning of Extraction Patterns from Unannotated corpora:Position statement” ? Manning, C. [1993], ``Automatic acquisition of a large subcategorization dictionary from corpora,'' Proceedings of the 31st Annual Meeting of the ACL. ? Sangal and Bansal, “A Modular Scalable Architecture for Text Processing : from Text to Trees to Text using XML” 《外贸英语函电》课程总结 《外贸英语函电》课程是国际贸易专业的必修课程，本课程系统讲述了外贸英语函电中常用文体的基本知识，并详细讲解了外贸业务磋商过程中各个环节往来函电的实例。本人从事该课程的教学三年，现做课程总结。 一、教学目标 通过外贸英语函电的学习，应当让学生达到以下要求：（1）通过阅读和学习介绍外贸交易业务过程中的章节，了解并熟悉外贸进出口业务的关键步骤；（2）重点掌握外贸业务信函的撰写要点，初步做到能再一般情况下写出内容确切、表达得体、符合规范、语句通顺、没有语法错误的信函（由于现代化的信息传送方式如传真的撰写和普通信函相似，故以掌握信函撰写为重点）；（3）了解外贸业务过程中的相关术语，包括交易条件和环境。（4）让学生了解业务信函的趋向，并学习电传、传真、邮件的有关知识及写法 二、教学方法 本课程采用多媒体教学，根据国贸专业学生英语的实际情况，采用不同的方法，提高教学质量，保证教学目标的实现。在教学过程中，我先后采用学生朗读、分组翻译、角色扮演等方法。同时，为了帮助学生复习和巩固所学知识，提高学生们的学习兴趣，定期使用经典欧美歌曲、经典电影片段等影像资料。 三、教学过程 基于“学生为本”的思想和理念，本课程的教学过程设计为：单词句式教学、函电教学、集中练习。单词句式部分，首先，跟读单词和词汇，并让同学们课后进行熟读熟记；其次，课堂听写，检验学生们的记忆情况；最后，结合经典英语歌曲，对重点词句进行强化，同时提高学生们的学习兴趣，营造轻松的学习氛围。在函电教学阶段，采用分组学习与教师重点讲解结合的方法，提高学生学习的自主性。提前布置函电，每组完成朗读、翻译两项任务，然后，对于学生朗读和翻译中存在的问题，进行及时纠正，并对重点用语和词汇进行讲解。集中练习阶段，主要是书写函电和完成课后习题。 四、学情分析 10国贸三个班级，分两班教学，即：10国贸1班和10国贸2、3班。总体上来看，大部分同学都能按照老师的教学方法和过程进行学习，效果也不错。10国贸1班情况好于其他班级，这个班级同学英语基础好，好学上进。国贸2班和3班基本上能按照课堂要求进行，总体还可以，但是部分同学记不住单词和用语，这可能与学习用功程度有关。三个班级同学课后作业都能按时完成，质量可以得到保证。在教学过程中，角色扮演环节，每个班级每组同学都能按照要求进行，用英语进行对话。应该说，教学过程都很顺利，教学目标基本达到。 五、考试分析 上学期，在外贸英语函电课程考试中，10国贸1班通过率87.5%，10国贸2班76%，10国贸3班的通过率率是66%。总体来看，外贸英语函电教学达到了教学目标的要求，国 L.1 1．五大基本句型2．零冠词精讲背诵 L.2 1.现在进行时与一般现在时2．感叹句的基本结构精讲 L.3 1.“给予”动词带双宾精讲背诵 L.4 1．现在完成时2．Accept vs.receive3.excited vs. exciting知识点和第五课重复，可以不讲 L.5 1.一般现在时与过去完成时的区别2．In 3 minutes vs. in 3 minutes’time3. 带way的短语精讲课文：背诵课文 L.6 1.in vs.on2.不定冠词的一般用法3 动词词组搭配精讲课文：背诵 L.7 1.expect的用法2．过去进行时3．When, while and as4. 小品词（副词vs介词）精讲 L.8 1.形容词和副词的比较级和最高级2．谓语动词的单复数确定精讲：背诵 L.9 1.基本时间介词的用法：at, in, on, during, through, till, until2 时间表示法次精讲 L.10 1.被动语态2．名词所有格，双重所有格（37）3.made in, made of, made from, made by精讲;鼓励背诵，加强语感 L.11 1.deserve的用法2．不定式作宾语动词后是否需要先加一个名词或代词次精讲：适合背诵L.12 1．一般将来时可以不讲 L.13 1.将来进行时和一般将来时的区别精讲 L.14 1．Except, except for, apart from次精讲 L.15 1.afford的用法2.interrupt的用法3．直接引语与间接引语精讲：背诵 L.16 1．Remind的用法2．Fail的用法3．if条件句精讲：背诵 L.17 1.in spite of2.介词的用法in, 3 情态动词精讲：鼓励背诵L. 18 1.have的多种用法（助动词，完全动词）2．关于give的词组自学课文L. 19 1.hurry的用法2．Can vs may; can vs could; may vs might3. might as well次精讲：设置场景，组对背诵 L.20 1.动名词充当主语和宾语2．Instead of vs. instead精讲：背诵 L.21 1.含助动词的被动语态2．Come into3.drive的不同用法4．Home vs. house次精讲 L.22 1 课后介词搭配练习文章次精讲 L.23 1.there is vs. it is自学课文 英语语法知识汇总-五年级 英语语法知识汇总 一、词类： 1、名词 这里强调两点：不可数名词都默认为单数，所以总是用is 名词复数如何加后缀： 2、人称代词和物主代词 人称代词：有主格和宾格之分。一般动词前用主格，动词后用宾格。 3、指示代词 4、冠词 有a、an、the。a和an的区别：an用于元音音素（一般就是元音字母a、e、i、o、u）前，a用于辅音音素前。 二、否定句： be动词（am、is、are）+not、 情态动词can+ not、 助动词（do、does）+ not 如何将一个肯定的陈述句改为否定句： 1、看句中有无be动词，如有，直接在be动词后+ not。 2、看句中有无情态动词，如有，直接在情态动词后+ not。 3、如上述二者都没有，就应用助动词+ not。分四个步骤： （1）肯定陈述句中本来是没有助动词的，要加上去，位置在主语（某人或某物）后，动词前。 （2）确定助动词用do、does，根据句中动词，动词是原形的助动词就用d o，动词是第三人称单数的助动词就用does， （3）在助动词后加not。 （4）原句中动词假如发生变化就要恢复成原形。 强调一点，有some的要考虑是否要用any。 三、一般疑问句。 如何将一个肯定的陈述句改为一般疑问句： 1、看句中有无be动词，如有，把be动词提到句首即可。 2、看句中有无情态动词，如有，把情态动词提到句首即可。 3、如上述二者都没有，就应把助动提到句首。分四个步骤： （1）肯定陈述句中本来是没有助动词的，要加上去，位置在主语（某人或某物）后，动词前。 I do like you. （2）确定助动词用do还是does，根据句中动词，动词是原形的助动词就用do，动词是第三人称单数的助动词就用does He does like you. （3）把助动词后提到句首。 Does he like you? （4）原句中动词假如发生变化就要恢复成原形。 强调一点，有some的要考虑是否要用any。 四、特殊疑问句。 表示疑问，有疑问词（在开头），回答有很多种可能。 1、Telex released b/l 电放提单 2、for your reference 做为你的参考 3、sign back 签回 4、Pls see attachment 请看附件 5、Pls sign back the telex releae letter . 请签回电放信 6、Pls find attached SO for your reference (yr ref.) 请查收附件做为你的参考 7、Pls make sure your cargo will send to our warehouse before tomorrow afternoon(18:00pm). 请尽量将你的货在明天下午6点前送到我们的仓库 8、Pls see attchment and cfm the b/l of XXX. 请收附件并且确认提单为XXX是否OK 9、I receipted the telex released requirment , but no payment receipt , we can’t telex release this mail it to me hurry! 我有收到电放函，但未收到付款收据，请尽快Email过来，否则我们不能电放此票货物。 10、Any further question, pls feel free to contact me, tks! 有任何问题，请及时联络我 11、Enclosed booking form for yr ref, pls fill it out by return 附件所涵内容给你参考，请填好后回传给我 新概念英语第二册语法知识点：主谓一致 I.单数主语： 1.当 every-, some-, any-, no- 等构成的不定代词及 each, either, neither 作主语时，谓语动词为单数。 ● Someone has told me about it. ● Neither of us likes the film. 2.当 every 或 each 位于两个由连接词连接的单数主语前，谓语动词用单数。(必记之要点!) ● Each girl and boy has a nickname. ● Every man and woman is welcome. 3.位于主谓之间的介词短语不影响谓语动词的数，谓语动词须和主语一致。常见介词短语有：in addition to, together with, as well as, along with, but, except, etc. (这是一几乎所有考试都不愿放过的重点。) ● The teacher along with his studen ts is going to the party. ● His parents as well as his elder sister have come to see him. II.复数主语： 1.当主语由 and 或 both ... and 连接，通常采用复数谓语动词。 ● Both the girl and the boy are his friends. 2.如主语是both, few或主语前有both, few, several, many 等限定词时，谓语动词须用复数。 ● Several novels have been written by her. ● Both got the news at the same time. 3.下列名词总是复数形式且谓语动词必须用复数。 trousers, pants, jeans, glasses, scissors, tweezers, plier, scales , compasses, etc. ● The trousers here are mine. 别忘了：如果以上名词受“a pair of” “the pair of”修饰时，谓语动词用单数。 The pair of pants is too dirty. III.单、复数的灵活使用： 1.当主语由 neither... nor, either... or, not only ... but also 或 or 连接时，谓语动词与nor, or, but also后面的词一致，在英语语法中，这被称之为“就近原则”。 ● Neither he nor I am going to the airport. ● Not only Mary but also her parents have gone abroad. 2.当主语前面有 none, all, some, any, most, half, majority等词时，谓语动词则根据主语的具体情况而采用相对应的单、复数。 ● All of the money has been in the bank. (money 是不可 数名词) ● All of us have been here. 《高中英语语法大全》(word下载版) 本文件内容丰富，讲解详细，层次分明，重点突出，包括高中英语中非常详尽的知识点、易错点、易混点、常考点等，是高中学生和老师非常实用而而且管用语法大全。适合不同层次的高中学生使用。 《高中英语语法大全》第01章名词 一、概说 名词是表示人、事物、抽象概念等名称的词，如boy 男孩，mother 母亲，news 消息，progress 进步，computer 计算机，Tom 汤姆，Paris 巴黎，Japan 日本，furniture 家具，等。 名词根据其词汇意义，通常分为专有名词和普通名词。专有名词主要指人、地方、组织、机构等的专有的名称，专有名词的第一个字母通常大写，如Mary 玛丽，Mr Green 格林先生，Beijing 北京，等；普通名词通常指人、物、概念等的一般名称。根据普通名词的语法性质，它又可以细为个体名词、物质名词、集合名词和抽象名词四类：个体名词表示人或物的个体，如girl 女孩，pen 钢笔，等；物质名词表示无法分为个体的实物，如wood 木头，meat 肉，等；集合名词表示若干个体组成的集合体，如：family 家庭，crowd 人群，等；抽象名词表示性质、行为、状态、感情等抽象概念，如work 工作，happiness 幸福，等。 二、名词的数 1.名词复数的构成方法 (1)在一般情况下，加词尾 -s： book / books 书 pen / pens 钢笔 face / faces 脸 (2)以 s, x, z, sh, ch 等结尾的名词，通常加词尾 -es： bus / buses 公共汽车 box / boxes 盒子 dish / dishes 盘子 注：有些以 ch 结尾的名词，由于其发音不是 [k] 而是 [tf]，那么其复数形式应加词尾–s，如stomach / stomachs 胃。 (3)以y 结尾的名词，其复数构成要分两种情况：以―辅音字母+y‖结尾的名词，将 y 改为 ies；以―元音字母＋y‖结尾的名词，直接加词尾s： city / cities 城市 boy / boys 男孩 key / keys 钥匙 本文档由实惠网（https://www.doczj.com/doc/0d5670728.html,）编制，版权所有，尽供外贸交流学习商业目的请联系实惠网（https://www.doczj.com/doc/0d5670728.html,） 外贸函电常用范文Set new business relationship 建立贸易关系，可以通过多种途径，比如通过驻外机构、国外商会、同业商行、银行、出国访问、商品交易会、报纸广告、市场调查等等。建立贸易关系的信函，要写得诚恳、热情、礼貌、得体，并将写信人的意图清楚地叙述完整，给对方留下深刻地印象，使其愿意与你交往。 Dear Mr. Jones: We understand from your information posted on https://www.doczj.com/doc/0d5670728.html, that you are in the market for textiles. We 本文档由实惠网（https://www.doczj.com/doc/0d5670728.html,）编制，版权所有，尽供外贸交流学习商业目的请联系实惠网（https://www.doczj.com/doc/0d5670728.html,） would like to take this opportunity to introduce our company and products, with the hope that we may work with Bright Ideas Imports in the future. We are a joint venture specializing in the manufacture and export of textiles. We have enclosed our catalog, which introduces our company in detail and covers the main products we supply at present. You may also visit our website https://www.doczj.com/doc/0d5670728.html,,which includes our latest product line. Should any of these items be of interest to you, please let us know. We will be happy to give you a quotation upon receipt of your detailed requirements. We look forward to receiving your enquires soon. Sincerely, John Roberts 本文档由实惠网（https://www.doczj.com/doc/0d5670728.html,）编制，版权所有，尽供外贸交流学习商业目的请联系实惠网（https://www.doczj.com/doc/0d5670728.html,） 《高中英语语法大全》 本文件内容丰富，讲解详细，层次分明，重点突出，包括高中英语中非常详尽的知识点、易错点、易混点、常考点等，是高中学生和老师非常实用而而且管用语法大全。适合不同层次的高中学生使用。 《高中英语语法大全》第01章名词 一、概说 名词是表示人、事物、抽象概念等名称的词，如boy 男孩，mother 母亲，news 消息，progress 进步，computer 计算机，Tom 汤姆，Paris 巴黎，Japan 日本，furniture 家具，等。 名词根据其词汇意义，通常分为专有名词和普通名词。专有名词主要指人、地方、组织、机构等的专有的名称，专有名词的第一个字母通常大写，如Mary 玛丽，Mr Green 格林先生，Beijing 北京，等；普通名词通常指人、物、概念等的一般名称。根据普通名词的语法性质，它又可以细为个体名词、物质名词、集合名词和抽象名词四类：个体名词表示人或物的个体，如girl 女孩，pen 钢笔，等；物质名词表示无法分为个体的实物，如wood 木头，meat 肉，等；集合名词表示若干个体组成的集合体，如：family 家庭，crowd 人群，等；抽象名词表示性质、行为、状态、感情等抽象概念，如work 工作，happiness 幸福，等。 二、名词的数 1.名词复数的构成方法 (1)在一般情况下，加词尾 -s： book / books 书 pen / pens 钢笔 face / faces 脸 (2)以 s, x, z, sh, ch 等结尾的名词，通常加词尾 -es： bus / buses 公共汽车 box / boxes 盒子 dish / dishes 盘子 注：有些以 ch 结尾的名词，由于其发音不是 [k] 而是 [tf]，那么其复数形式应加词尾–s，如stomach / stomachs 胃。 (3)以y 结尾的名词，其复数构成要分两种情况：以“辅音字母+y”结尾的名词，将 y 改为 ies；以“元音字母＋y”结尾的名词，直接加词尾s： city / cities 城市 boy / boys 男孩 key / keys 钥匙 福建省高等教育自学考试应用心理学专业（独立本科段） 《心理学研究方法》课程考试大纲 第一部分课程性质与目标《心理学研究方法》是福建省高等教育自学考试应用心理学专业（独立本科段）的一门专业基础必修课程，目的在于帮助考生了解和掌握心理学研究的理论基础和主要方法，检验考生对心理学研究理论基础与主要方法，检验考生对心理学研究方法的基本知识和主要内容的掌握水平与应用能力。 心理学研究的对象是心理现象。它的研究主题十分广泛：即涉及人的心理也涉及动物的心理；即涉及个体的心理也涉及群体的心理；即涉及有意识的心理也涉及潜意识的心理；即涉及与生理过程密切相关的心理也涉及与社会文化密切相关的心理。心理学研究是一种以经验的方式对心理现象进行科学探究的活动。由于心理学的研究方法是以经验的或实证的资料为依据的，因而使心理学与哲学相区别，也与人文学科相区别。 设置本课程的具体目的要求是，学习和掌握心理学研究方法的基本理论和基本技能，将有助于学生们理解心理学的基本概念、基本原理和基本理论。理解心理学家在探索心理与行动时所做的一切，有助于考生将来为心理学的发展做出有益的贡献。 第二部分考核内容与考核目标 第一编心理学研究基础 第一章心理学与科学 一、学习目的与要求 通过本章学习，要求考生了解心理学的性质，了解心理学科学研究的方法、特征及基本步骤，理解心理学研究的伦理问题和伦理规范。 二、考核知识点与考核目标 1、识记： （1）心理学的含义； （2）心理学科学研究的特征：系统性、重复性、可证伪性和开放性； （3）知情同意。 2、领会： （1）一般人探索世界的常用方法； （2）心理学研究主要包含哪几个步骤； （3）科学研究的开放性主要表现在哪几方面； 3、应用： （1）根据科学研究的特征来分析某些心理学的研究； （2）心理学研究的伦理问题及以人为被试的研究的伦理规范来分析是否可以在心理学研究中使用欺骗的方法。 (1)贵函YourletterYourfaouryouresteemedletterYouresteemedfaourYouraluedletterYouraluedfaourYournote YourcommunicationYourgreatlyesteemedletterYoureryfriendlynoteYourfriendlyadiceYours. (2)本信，本函Our(my)letterOur(my)respectsOurs(mine)ThislettertheselinesThepresent. (3)前函ThelastletterThelastmailThelastpostthelastcommunicationThelastrespects(自己的信)Thelastfaour(来信) (4)次函ThenextletterThenextmailThenextcommunicationTheletterfollowingthefollowing. (5)贵函发出日期Yourletterof(the)5thMayYourfaourdated(the)5thJuneYoursofthe3rdJulyYoursunderdate(of)the5thJ ulyYourletterbearingdate5thJulyYourfaourofeendate(AE)YourletterofyesterdayYourfaourofyesterd ay''sdateYourletterdatedyesterday. (6)贵方来电、电传及传真YourtelegramYourwireYourcablegram(从国外)Yourcodedwire(密码电报)YourcodemessageYourciphertelegramYourwirelesstelegramYourTELEXYourFax. (7)贵方电话YourtelephonemessageYourphonemessageYourtelephoniccommunicationYourtelephonecallYourri ng. (8)通知(Noun)AdiceNoticeInformationNotificationCommunicationAreportNewsIntelligenceMessage.(erb )(通知，告知)Tocommunicate(afact)toToreport(afact)to...onToapprise(aperson)ofTolet(aperson)knowToacqu aint(aperson)withTointimate(afact)toTosendwordtosendamessageTomailanoticetowrite(aperson)in formationTogienotice(预告)Tobreakanewsto(通知坏消息)Toannounce(宣布). (9)回信(Noun)AnanswerAreplyAresponse.(erb)ToanswerToreplyTogieareplyTogieone''sanswer TomakeananswerTosendananswerTowriteinreplyToanswerone''sletter.(特此回信)ReplytoAnsweringtoInanswertoInreplytoInresponseto.(等候回信)ToawaitananswerTowaitforananswer.(收到回信)TogetananswerTofaouronewithananswerTogetaletteranswered. (10)收讫，收到(Noun)Receipt(收到)Areceipt(收据)Areceier(领取人，取款人)Arecipient(收款 人)(erb)ToreceieTobeinreceiptofTobeto(at)handTocometohandTobeinpossesionofTobefaouredwith TogetTohaeTohaebefore(aperson)Tomakeoutareceipt(开出收据)Toacknowledgereceipt(告知收讫). (11)确认ToconfirmConfirmingConfirmationInConfirmationof(为确认...,为证 —－可编辑修改，可打印—— 别找了你想要的都有！ 精品教育资料——全册教案，，试卷，教学课件，教学设计等一站式服务—— 全力满足教学需求，真实规划教学环节 最新全面教学资源，打造完美教学模式 新概念英语第一册语法点梳理 新概念一共144课，其中单课为课文，双课为语法和练习。整本书是以单数课为正课，并附带有插图而双数课则是针对单数课所讲的内容有针对性地进行练习，从此出展现出整个新概念一教材区别于其他教材的独特之处。以下是对新概念一整本教材的理解和剖析，以供各位对整个课本的理解和把握上参考和借鉴。 首先根据课本中出现的时态来分析： 本册书的语法出现层次性和规律性是很强的，首先我们先来整本书中都出了哪些时态，这些时态的具体分布和讲解时我们大家需要注意的递进性。 Lesson 31—34 现在进行时Lesson 37—40 第一次出现be going to 的将来时Lesson 51—56 一般现在时Lesson 67—76 为一般过去式 Lesson 83—90 为现在完成时Lesson 91—96 为一般将来时(will) Lesson 117—118 过去进行时Lesson 119—120 过去完成时 除去前面所有时态和句型所占据的76课我们一起来看一下以下的68课，每一课小的语言点，语法点都是在什么地方，应该用什么样的方式来讲解。新概念一的每一个单课的重点都是出现双课的标题和课后的练习题里面。 Lesson1—2 语言点：与陌生人说话或引起别人的注意。Excuse me. Yes? Pardon? Thank you very much. 语法点：主系表结构this为主语，名词做表语1的一般疑问句以及它的肯定回答。Is this your handbag? Yes, it is. Lesson 5—6 语言点：如何介绍别人。This is Miss Sophie Dupont. Nice to meet you. 语法点：主语为第三人称单数的主系表结构。She is French. He is German. It’s a Volvo.(L6) a/an 的使用。 Lesson 7—8 语言点：如何自我介绍和相互认识。 语法点：主语为第二人称的主系表结构。Are you French? What nationality are you? What’s your job? 特殊疑问句。 Lesson 9—10 语言点：朋友或熟识的人之间如何相互问候。How are you? 语法点：主系表结构形容词做表语。 介词短语表位置near the window, on the television, on the wall Lesson 29—30 语言点：如何发号命令。语法点：祈使句（肯定）;动词与宾语的固定搭配。 Lesson 37—38 语言点：如何表达将要做的事情。 语法点：现在进行时态be going to do结构表达将要发生的事情。There be 句型的一般疑问句形式。 英语语法大全 一、词类、句子成分和构词法： 1、词类：英语词类分十种： 名词、形容词、代词、数词、冠词、动词、副词、介词、连词、感叹词。1、名词(n.)：表示人、事物、地点或抽象概念的名称。如：boy, morning, bag, ball, class, orange. 2、代词(pron.)：主要用来代替名词。如：who, she, you, it . 3、形容词(adj..)：表示人或事物的性质或特征。如：good, right, white, orange . 4、数词(num.)：表示数目或事物的顺序。如：one, two, three, first, second, third, fourth. 5、动词(v.)：表示动作或状态。如：am, is,are,have,see . 6、副词(adv.)：修饰动词、形容词或其他副词,说明时间、地点、程度等。 如：now, very, here, often, quietly, slowly. 7、冠词(art..)：用在名词前,帮助说明名词。如：a, an, the. 8、介词(prep.)：表示它后面的名词或代词与其他句子成分的关系。如in, on, from, above, behind. 9、连词(conj.)：用来连接词、短语或句子。如and, but, before . 10、感叹词(interj..)表示喜、怒、哀、乐等感情。如：oh, well, hi, hello. 2、句子成分：英语句子成分分为七种：主语、谓语、宾语、定语、状语、表语、宾语补足语。 1、主语是句子所要说的人或事物,回答是“谁”或者“什么”。通常用名词或 代词担任。如：I’m Miss Green.(我是格林小姐) 2、谓语动词说明主语的动作或状态,回答“做(什么)”。主要由动词担任。如： Jack cleans the room every day. (杰克每天打扫房间) 3、表语在系动词之后,说明主语的身份或特征,回答是“什么”或者“怎么样”。 通常由名词、代词或形容词担任。如：My name is Ping ping .(我 的名字叫萍萍) 4、宾语表示及物动词的对象或结果,回答做的是“什么”。通常由名词或代词 外贸函电词汇 UNIT ONE 1 ESP： English for Specific/Special Purposes 专门用途英语 2 BE： Business English 商务英语 3 Correspondence for Import & Export 进出口函电 4 Good quality stationery 优质信笺信纸 5 Neat typing 整洁 6 Even spacing 间隔匀称 7 Short paragraphs 段落精短 8 Correct Grammar， Spelling and Punctuation 语法、拼写、标点正确UNIT TWO INQUIRIES 1 potential business 潜在业务 2 prospective customer 潜在顾客 3 customers of long standing 长期顾客 4 potential supplier 潜在供应商 5 trade fair 贸易博览会 6 the latest issue of 最新一期… 7 integrated software package 完整软件包 8 substantial order 大宗订单 9 quantity discount 数量折扣 10 cash discount 现金折扣 11 list price 标价、目录价格 12 export terms 出口条件 13 pictured/illustrated catalog 带插图的商品目录 14 article number / Art. No. 货号 15 bulk buyer 大买户 16 business concern 商行 17 business relations / relationship 业务关系 18 business status 业务状况 19 commercial counselor 商务参赞 20 commercial counselor's office 商务参赞处 21 means of packing 包装方法 22 parent company 母公司 23 sales literature 促销资料 24 trading association 贸易关系 25 trade journal 行业刊物 26 firm offer 实盘 27 non-firm offer： offer without engagement 虚盘 28 trade discount 同业/批发折扣 29 bill of exchange （bill / draft； B/E）汇票 30 documents against payment： D/P 付款交单 31 shipping documents 装运单据 32 line of business 业务/经营范围 33 specific inquiry 具体询价 UNIT THREE REPLIES AND QUOTATIONS 1 regular customer 老顾客 新概念二册语法总结-标准化文件发布号：（9556-EUATWK-MWUB-WUNN-INNUL-DDQTY-KII 新概念二册语法总结（新东方） 1. 简单句的结构: 主语+谓语+宾语+状语（地点状语+方式状语+时间状语）时间状语还可以放在句首 2. 一般现在时，现在进行时感叹句：what+名词+主语+谓语， how+形容词/副词+主语+谓语 频度副词：位于主语和谓语之间，常见频度副词及其程度的深浅： always, often, usually, frequently, sometimes, hardly, rarely, seldom, never I always buy CDs on Sundays. 3. 一般过去时直接宾语和间接宾语： 主语＋及物动词＋间接宾语＋直接宾语直接宾语是及物动词的直接对象，间接宾语是及物动词的动作所涉及的人或事务，也可以说间接宾语表示动作是对谁做的，或者是为谁做的。所以间接宾语要用名词或者宾格代词来担当。 He gives me a book. me间接宾语, a book直接宾语 直接宾语和间接宾语的位置调换时要加一个介词to或for，to表示动作对谁而做， for表示动作为谁而做。 主语＋及物动词＋直接宾语＋介词+间接宾语Give the book to me. Send a letter to him. I bought a coat for my mother. 4. 现在完成时，注意频度及时间副词的位置 receive/take 5. 复习：一般过去时与现在完成时的区别in the way/on the way/in this way/by the way/in a way/get out of my way/Don’t stand in my way./by the way/no way spare/to spare 6. 冠词用法（一）1．不定冠词用于修饰单数可数名词，当一个单词的第一个发音为元音时要用冠词an, 如果是辅音用a即可。 2．不定冠词还可以用来修饰一类事物，有时候也可以省略： A cat is a lovely animal. Cat is a lovely animal. 3．不可数名词加冠词表示一类事物：Apple is a fruit. 4．如果特制某人,某物或上文提过的人或物时要用定冠词the 5． Some表示一些，可以修饰可数名词及不可数名词。 6．在姓名，城市，地名，过名，月份，节日，星期前不加冠词，季节前也一般不加，特指时加。 put on/take off/look for/look after/knock at/knock sth. off/knock off/knock %off/knock over/knock out 7. 过去进行时，时间状语短语：在短语中动词后面的介词或副词叫小品词，有些小品词既可以紧跟在动词后面也可以跟在动词的宾语后面，而有些则只能紧跟着动词使用。 8. 形容词的比较级与最高级单音节词的比较级最高级：直接加-er, -est: small-smaller, smallest 以e结尾的加-r, -st: large-larger, largest 以辅音字母加y结尾的词变y为i加-er, -est: busy-busier, busiest 个人整理切勿外传 按：本套资料省去了名词、代词、形容词等部分，保留了最最核心的句型和动词。希望能有所帮助。 第一章名词性从句 在句子中起名词作用的句子叫名词性从句。名词性从句的功能相当于名词词组, 它在复合句中能担任主语、宾语、表语、同位语、介词宾语等，因此根据它在句中不同的语法功能，名词从句又可分别称为主语从句、宾语从句、表语从句和同位语从句。 一．主语从句 主语从句是在复合句中充当主语的从句，通常放在主句谓语动词之前或由形式主语it代替，而本身放在句子末尾。 1. It作形式主语和it引导强调句的比较 It作形式主语代替主语从句，主要是为了平衡句子结构，主语从句的连接词没有变化。而it引导的强调句则是对句子某一部分进行强调，无论强调的是什么成分，都可用连词that。被强调部分指人时也可用who/whom。例如： a) It is a pity that you didn't go to see the film. 你不去看那场电影真可惜。 b) It doesn't interest me whether you succeed or not.我对你成功与否不感兴趣。 c) It is in the morning that the murder took place.谋杀案是在早上发生的。（强调句型） d) It is John that broke the window.是John打碎的窗户。（强调句型） 2. 用it作形式主语的结构 (1) It is ＋名词＋从句 It is a fact that …; It is an honor that…; It is common knowledge that… (2) It is ＋形容词＋从句 It is natural that…It is strange that… (3) It is ＋不及物动词＋从句 It seems that…It happened that…It appears that… (4) It ＋过去分词＋从句 It is reported that…It has been proved that…It is said that… 3. 主语从句不可位于句首的五种情况： （1）if引导的主语从句不可居于复合句句首。 （2）It is said /reported…结构中的主语从句不可提前。例如： 正确表达：It is said that President Jiang will visit our school next week. 错误表达：That President Jiang will visit our school next week is said. （3）It happens/occurs…结构中的主语从句不可提前。例如： 正确表达：It occurred to him that he failed in the examination. 错误表达：That he failed in the examination occurred to him. （4）It doesn't matter how/whether …结构中的主语从句不可提前。例如： 正确表达：It doesn't matter whether he is wrong or not. 错误表达：Whether he is wrong or not doesn't matter. （5）含主语从句的复合句是疑问句时，主语从句不可提前。例如： 正确表达：Is it likely that it will rain in the evening? 2、 for your reference 做为你的参考 3、 sign back 签回 4、 Pls see attachment 请看附件 5、 Pls sign back the telex releae letter . 请签回电放信 6、 Pls find attached SO for your reference (yr ref.) 请查收附件做为你的参考 7、 Pls make sure your cargo will send to our warehouse before tomorrow afternoon(18:00pm). 请尽量将你的货在明天下午6点前送到我们的仓库 8、 Pls see attchment and cfm the b/l of XXX. 请收附件并且确认提单为XXX是否OK？ 9、 I receipted the telex released requirment , but no payment receipt , we can’t telex release this mail it to me hurry! 我有收到电放函，但未收到付款收据，请尽快Email过来，否则我们不能电放此票货物。 10、Any further question, pls feel free to contact me, tks! 有任何问题，请及时联络我 11、Enclosed booking form for yr ref, pls fill it out by return 附件所涵内容给你参考，请填好后回传给我 12、Enclosed our airport warehouse address for yr send the cargo tomorrow morning 请于明天早上送货，附件为飞机场仓库地址 13、Pls find the attached , 1 filed for two TELEX RELEASE B/L to CHICAGO and check it ‘s OK or NOT 请收附件，我已电放好了去芝加哥的两票货物，请检查是否OK？ 14、Have received 已有收到 15、Received with tks. 收到，谢谢 16、Received and will arrange 已收到且将会安排 17、The s/o can use till next week. 这份S/O下星期也可以使用 18、Please confirm the attached bill by 请确认附件并传回 19、Please kindly advice yr PO#. 请提供PO号码 20、Please find the attached invoice for yr 请查收附件中的付款单做参考 21、I have release S/O to you , Pls received it . 我已有释放S/O给你，请查收。 22、Pls find attached the application form for telex release. 请查收附件中的电放申请函 23、Pls see the attached that we had arranged the hb/l telex release. Here is the copy for you in your information.外贸英语函电课程总结
最新新概念英语第二册语法总结
【五年级】英语语法知识汇总(全)
外贸函电常用回复
新概念英语第二册语法知识点：主谓一致
《高中英语语法汇总》
外贸函电常用范文
《高中英语语法汇总》最全、最实用
(完整版)心理学研究方法
外贸函电常用语法汇总
新概念英语一册语法点梳理
高中英语语法大全总结
外贸函电词汇
新概念二册语法总结
完整高中英语语法最重要知识汇总
外贸函电常用回复
相关主题
文本预览