CN106354712A - Method and system of expressing semantics of verbs based on concept of parameters - Google Patents

Abstract

The present invention provides a method and system for expressing verb semantics based on the concept of parameters, comprising the following steps: step 1: extracting the dependency relationship between verb-object and verb-subject; step 2: according to the dependency relationship between verb-object and verb-subject Calculate the entropy value of the verb parameter to the pattern, and calculate the mutual information value of the verb parameter to the verb, and calculate the weight of the verb parameter; Step 3: Conceptualize the verb parameter, that is, find the k group with the largest weight of the verb parameter. The present invention creatively uses an external knowledge base to represent the semantics of verbs, and provides users with parameters to select the semantic granularity of verbs, so that the obtained semantic concepts of verbs are of moderate size, which can be read by humans or directly calculated by machines .

Description

Method and system for expressing verb semantics based on parameter concept

technical field

The present invention relates to natural language processing in the field of computer technology, in particular to a method and system for expressing verb semantics based on parameter concepts.

Background technique

With the deepening of the application of artificial intelligence technology, artificial intelligence technology plays more roles in human life. Among them, natural language understanding technology is an important technology for computers to understand human language, and the most difficult part is to understand the semantics of human language.

Verbs play a central role in understanding the syntax and grammar of sentences. The distributional assumption ( ) states that the semantics of a word can be represented by its contextual information, such as the words around it. A verb has its unique role in a sentence, because it contains the dependent relationship with its subject and object, so the semantics of this verb can be expressed with the subject and object of this verb. Similar systems include ReVerb(), which uses a "bag of words" approach, but this model has the following disadvantages:

1) The relationship of synonyms cannot be considered;

2) The "bag of words model" has a high dimensionality, so the calculation efficiency is low;

3) The generated model is not human readable.

In order to make up for these deficiencies, a very natural way is to use abstract concepts or types of these subjects and objects to represent, rather than using these words directly. Similar systems include FrameNet(), which expresses the semantics of verbs by humans annotating their subject and object types. However, this system also has some obvious flaws:

1) The workload of manual labeling is huge and cannot be expanded;

2) The abstraction of verb parameters is too high. For example, the object concept of the verb "eat" is only "Ingestibles", so it cannot express the multiple semantics of the verb.

After retrieval, the application number is: 201010290860.5, and the name is: a method for extracting verb semantic information based on event ontology. In this invention, a method for extracting verb semantic information based on event ontology is used. By matching verbs and verb roles, the method improves The accuracy of recognizing verbs. However, this invention does not generate a human-readable and machine-computable concept set for verbs, and cannot change the semantic granularity of verbs set by humans.

The application number is: 200510088741.0, and the title is: A semantic analysis method for dissolving verb ambiguity structure in sentence analysis, in which the semantics of verb ambiguity structure dissolving in sentence analysis is used; including: construction of semantic model and semantic analysis In the method, the semantic model is used to express the ambiguity structure of the verb, and the semantic analysis method realizes the judgment, resolution and calculation of the ambiguity structure according to the semantic model. This invention establishes a unified semantic model for expressing the ambiguous structure of verbs, and raises the ambiguous structure to the sentence level for processing; however, no external knowledge base is used, so the semantics of verbs cannot be expressed.

To sum up, the ReVerb system expresses verbs with too fine granularity, while FrameNet expresses verbs with too coarse granularity, so we urgently need an algorithm and system that can accurately express verb semantics.

Contents of the invention

In view of the defects in the prior art, the object of the present invention is to provide a method and system for expressing verb semantics based on the parameter concept.

The method for expressing verb semantics based on the parameter concept provided by the present invention comprises the following steps:

Step 1: Extract the verb and object, and the dependency relationship between the verb and the subject;

Step 2: Calculate the entropy value of the verb parameter for the pattern according to the dependency relationship between the verb and the object, and the verb and the subject, and calculate the mutual information value of the verb parameter for the verb, and calculate the weight of the verb parameter;

Step 3: Conceptualize the verb parameters, that is, find the k-cluster with the largest verb parameter weight.

Preferably, said step 2 includes:

Step 2.1: Calculate the entropy value of the verb parameter for the pattern, the larger the entropy value, the better the quality of the verb parameter, and the quality refers to the reliability of a certain word as the parameter of the verb;

${\mathrm{<span\; class="notranslate">\; Entropy</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

In the formula: Entropy _v (e) represents the mode entropy of the word e for the verb v, P(m) represents the probability of the occurrence of the mode m, m represents the mode m, that is, different combinations of the word e and the verb v, M _e,v represents All collocations of the word e and the verb v;

Step 2.2: Calculate the mutual information value of the verb parameter for the verb. The higher the mutual information value, the better the quality of the verb parameter. Specifically, the calculation formula of the binary mutual information used is as follows:

${\mathrm{<span\; class="notranslate">\; MI</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$

In the formula: MI _v (e) represents the mutual information of the word e relative to the verb v, p(v, e) represents the probability that the verb v and the word e appear together in the corpus, and p(v) represents the probability that the verb v appears, p(e) represents the probability of word e appearing;

Step 2.3: Calculate the weight value Q _v (e) of the verb parameter, the calculation formula is as follows:

Qv(e)= _Entropyv (e)× _{MIv (} e).

Preferably, the step 3 includes: using a branch-and-bound algorithm to find the k-cluster with the largest weight; wherein, the branch-and-bound algorithm refers to: constructing a search tree, each layer of the tree, except the root node, represents Is whether to select a parameter concept, where the left branch is to select the parameter concept, and the right branch indicates not to select the parameter concept; when the number of selected parameter concepts is k, judge whether the k parameter concepts are a group in the graph , if yes, return true; if not a clique, return false.

The system for expressing verb semantics based on the parameter concept provided by the present invention includes:

Dependency extraction module: used to extract verbs and objects, as well as the dependency relationship between verbs and subjects;

Verb parameter weight calculation module: used to calculate the entropy value of the verb parameter for the mode according to the dependency relationship between the verb and the object, the verb and the subject, and calculate the mutual information value of the verb parameter for the verb, and calculate the weight of the verb parameter;

The verb parameter conceptualization module is used to conceptualize the verb parameter, that is, to find the k-cluster with the largest verb parameter weight.

Preferably, the verb parameter weight calculation module includes:

Entropy value calculation module: calculate the entropy value of the verb parameter for the pattern, the larger the entropy value, the better the quality of the verb parameter, and the quality refers to the reliability of a certain word as the parameter of this verb; in the calculation formula, it is as follows:

${\mathrm{<span\; class="notranslate">\; Entropy</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

In the formula: Entropy _v (e) represents the mode entropy of the word e for the verb v, P(m) represents the probability of the occurrence of the mode m, m represents the mode m, that is, different combinations of the word e and the verb v, M _e,v represents All collocations of the word e and the verb v;

Mutual information value calculation module: Calculate the mutual information value of the verb parameter for the verb. The higher the mutual information value, the better the quality of the verb parameter. Specifically, the calculation formula for the binary mutual information used is as follows:

${\mathrm{<span\; class="notranslate">\; MI</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$

In the formula: MI _v (e) represents the mutual information of the word e relative to the verb v, p(v, e) represents the probability that the verb v and the word e appear together in the corpus, and p(v) represents the probability that the verb v appears, p(e) represents the probability of word e appearing;

The weight value calculation module of the verb parameter: calculate the weight value Q _v (e) of the verb parameter, and the calculation formula is as follows:

Qv(e)= _Entropyv (e)× _{MIv (} e).

Preferably, the verb parameter conceptualization module uses a branch-and-bound algorithm to find the k-cluster with the largest weight; wherein, the branch-and-bound algorithm refers to: constructing a search tree, each layer of the tree, except the root node, represents Is whether to select a parameter concept, where the left branch is to select the parameter concept, and the right branch indicates not to select the parameter concept; when the number of selected parameter concepts is k, judge whether the k parameter concepts are a group in the graph , if yes, return true; if not a clique, return false.

Compared with the prior art, the present invention has the following beneficial effects:

1. The method for expressing verb semantics based on parameter concepts creatively utilizes an external knowledge base to express verb semantics, and provides parameters for users to select the semantic granularity of verbs, so that the obtained verb semantic concepts are of moderate size.

2. The method for expressing verb semantics based on parameter concepts provided by the present invention can obtain verb semantic concepts that can be read by humans, and can also be directly calculated by machines.

Description of drawings

Other characteristics, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

Fig. 1 is a technical program frame diagram of the present invention;

Figure 2 is a concept map structure;

Fig. 3 is another conceptual diagram structure corresponding to Fig. 2;

FIG. 4 is a schematic diagram of a branch-and-bound search tree.

detailed description

The present invention will be described in detail below in conjunction with specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

According to the method and system for expressing verb semantics based on the parameter concept provided by the present invention, an algorithm for accurately extracting the concept of verb parameters is first proposed, and the extracted verb parameters are all from a knowledge base, and the knowledge base includes Probase and WordNet. These knowledge bases store a large number of relationships between concepts and entities, which are called "IsA" relationships. For example, if an apple (apple) is a kind of (IsA) fruit, then the fruit is a concept and the apple is an entity.

The semantic overlap scores for two concepts are defined as follows:

$\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; v</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; \&cap;</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; ;</span>$

in, and represent the entity sets _of concepts c1 and _c2 , respectively. The present invention then converts the problem of verb parameter conceptualization into a k-clique problem of finding the maximum weight in an undirected graph. A concept map G=(C,L,W), where C represents the concept set of the knowledge base, L represents the semantic overlap score between two concepts is less than a certain threshold τ, W represents the weight of a certain concept in the graph, Used to indicate the quality of the concept relative to a verb, that is, whether the concept can express the usage of the verb, such as "food" relative to the verb "eat". Figure 2 and Figure 3 show the structure of a concept map.

The concept weight is defined by calculating the number of entities contained under the concept map, that is, assuming that the importance of these entities is consistent for a certain verb, but this situation is generally not satisfied. Therefore, the present invention defines that the quality of the entity e of a certain verb v is represented by the verb parameter weight Q _v (e), corresponding to step 2.3, so the verb parameter weight for the concept c of the verb v can be defined as follows:

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; A</span>}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \}</span>}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>$

In the formula: w _v (c) represents the parameter weight of concept c relative to verb v; e is a certain entity, c is a certain concept, and eIsA c represents that entity e is an instance of concept c, such as "apple" is "food An instance of ".

Therefore, the problem of verb parameter conceptualization is to find a k group (defined as C _k ) in the conceptual graph G=(C, L, W), so that the value of the following equation is the largest, and the equation is as follows:

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

Specifically, include the following steps:

Step 1: Extract the verb and object, and the dependency relationship between the verb and the subject;

Step 2: Calculate the entropy value of the verb parameter for the pattern according to the dependency relationship between the verb and the object, and the verb and the subject, and calculate the mutual information value of the verb parameter for the verb, and calculate the weight of the verb parameter;

Step 3: Conceptualize the verb parameters, that is, find the k-cluster with the largest verb parameter weight.

Said step 2 includes:

Step 2.1: Calculate the entropy value of the verb parameter for the pattern, the larger the entropy value, the better the quality of the verb parameter (that is, the reliability of a certain word as the parameter of the verb);

${\mathrm{<span\; class="notranslate">\; Entropy</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

In the formula: Entropy _v (e) represents the pattern entropy of word e for verb v, P(m) represents the probability of pattern m appearing, m represents pattern m (different combinations of word e and verb v), M _e,v represents All collocations of the word e and the verb v;

Step 2.2: Calculate the mutual information value of the verb parameter for the verb. The higher the mutual information value, the better the quality of the verb parameter. Specifically, the calculation formula of the binary mutual information used is as follows:

${\mathrm{<span\; class="notranslate">\; MI</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$

In the formula: MI _v (e) represents the mutual information of the word e relative to the verb v, p(v, e) represents the probability that the verb v and the word e appear together in the corpus, and p(v) represents the probability that the verb v appears, p(e) represents the probability of word e appearing;

Step 2.3: Calculate the weight value of the verb parameter:

Qv(e)= _Entropyv (e)× _{MIv (} e).

The step 3 includes: using a branch and bound algorithm to find the k-cluster with the largest weight.

In Figure 2, C0, C1, C2 and C3 respectively represent four parameter concepts, among which C0 and C3, C1 and C3 have a high degree of coincidence (relatively close in semantics), thus constructing a graph, in which C0 and C3, C1 There is no edge connection with C3, as shown in Figure 3. The goal is to find the k group with the largest weight in the graph shown in Figure 3, where k=3; the branch and bound algorithm is worth constructing a search tree, as shown in Figure 4, each layer of the tree (excluding the root Node) indicates whether to select a parameter concept, where the left branch is to select the parameter concept, and the right branch indicates not to select the parameter concept; when the number of selected parameter concepts is k, it is judged whether the k parameter concepts are in the graph is a clique, if it is, it will return correct, if it is not a clique, it will return an error, see Figure 4 for an example.

Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention.

Claims (6)

1. A method for expressing verb semantics based on parameter concept, is characterized in that, comprises the steps: Step 1: Extract the verb and object, and the dependency relationship between the verb and the subject; Step 2: Calculate the entropy value of the verb parameter for the pattern according to the dependency relationship between the verb and the object, and the verb and the subject, and calculate the mutual information value of the verb parameter for the verb, and calculate the weight of the verb parameter; Step 3: Conceptualize the verb parameters, that is, find the k-cluster with the largest verb parameter weight. 2. the method for expressing verb semantics based on parameter concept according to claim 1, is characterized in that, described step 2 comprises: Step 2.1: Calculate the entropy value of the verb parameter for the pattern, the larger the entropy value, the better the quality of the verb parameter, and the quality refers to the reliability of a certain word as the parameter of the verb; ${\mathrm{<span\; class="notranslate">\; Entropy</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$ In the formula: Entropy _v (e) represents the mode entropy of the word e for the verb v, P(m) represents the probability of the occurrence of the mode m, m represents the mode m, that is, different combinations of the word e and the verb v, M _e,v represents All collocations of the word e and the verb v; Step 2.2: Calculate the mutual information value of the verb parameter for the verb. The higher the mutual information value, the better the quality of the verb parameter. Specifically, the calculation formula of the binary mutual information used is as follows: ${\mathrm{<span\; class="notranslate">\; MI</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; g</span>}\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$ In the formula: MI _v (e) represents the mutual information of the word e relative to the verb v, p(v, e) represents the probability that the verb v and the word e appear together in the corpus, and p(v) represents the probability that the verb v appears, p(e) represents the probability of word e appearing; Step 2.3: Calculate the weight value Q _v (e) of the verb parameter, the calculation formula is as follows: Qv(e)= _Entropyv (e)× _{MIv (} e). 3. the method for expressing verb semantics based on parameter concept according to claim 1, is characterized in that, described step 3 comprises: adopt the algorithm of branch and bound to find the k group of maximum weight; Wherein, the algorithm of described branch and bound refers to : Construct a search tree. Each layer of the tree, except the root node, indicates whether to choose a certain parameter concept, where the left branch means to choose the parameter concept, and the right branch means not to choose the parameter concept; when the selected parameter When the number of concepts is k, judge whether the k parameter concepts are a clique in the graph, if yes, return correct; if not a clique, return error. 4. A system for expressing verb semantics based on parameter concepts, characterized in that it includes: Dependency extraction module: used to extract verbs and objects, as well as the dependency relationship between verbs and subjects; Verb parameter weight calculation module: used to calculate the entropy value of the verb parameter for the mode according to the dependency relationship between the verb and the object, the verb and the subject, and calculate the mutual information value of the verb parameter for the verb, and calculate the weight of the verb parameter; The verb parameter conceptualization module is used to conceptualize the verb parameter, that is, to find the k-cluster with the largest verb parameter weight. 5. the system based on parameter concept expression verb semantics according to claim 4, is characterized in that, described verb parameter weight calculation module comprises: Entropy value calculation module: calculate the entropy value of the verb parameter for the pattern, the larger the entropy value, the better the quality of the verb parameter, and the quality refers to the reliability of a certain word as the parameter of this verb; in the calculation formula, it is as follows: ${\mathrm{<span\; class="notranslate">\; Entropy</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$ In the formula: Entropy _v (e) represents the mode entropy of the word e for the verb v, P(m) represents the probability of the occurrence of the mode m, m represents the mode m, that is, different combinations of the word e and the verb v, M _e,v represents All collocations of the word e and the verb v; Mutual information value calculation module: Calculate the mutual information value of the verb parameter for the verb. The higher the mutual information value, the better the quality of the verb parameter. Specifically, the calculation formula for the binary mutual information used is as follows: ${\mathrm{<span\; class="notranslate">\; MI</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; g</span>}\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$ In the formula: MI _v (e) represents the mutual information of the word e relative to the verb v, p(v, e) represents the probability that the verb v and the word e appear together in the corpus, and p(v) represents the probability that the verb v appears, p(e) represents the probability of word e appearing; The weight value calculation module of the verb parameter: calculate the weight value Q _v (e) of the verb parameter, and the calculation formula is as follows: Qv(e)= _Entropyv (e)× _{MIv (} e). 6. The system for expressing verb semantics based on the parameter concept according to claim 4, wherein the verb parameter conceptualization module uses a branch and bound algorithm to find the k group of maximum weight; wherein, the branch and bound algorithm refers to : Construct a search tree. Each layer of the tree, except the root node, indicates whether to choose a certain parameter concept, where the left branch means to choose the parameter concept, and the right branch means not to choose the parameter concept; when the selected parameter When the number of concepts is k, judge whether the k parameter concepts are a clique in the graph, if yes, return correct; if not a clique, return error.