CN104090918B - Sentence similarity calculation method based on information amount - Google Patents

Abstract

The present invention relates to a kind of sentence similarity calculation method based on the amount of information, comprising the following steps: first, determine the meaning of the word through the concept with the largest amount of information between two sentence words; then use the hierarchical structure of the semantic web and corpus statistics to Calculate the amount of information of words and the amount of public information between multiple words; then apply the principle of inclusion and exclusion in combinatorics to calculate the total information amount of multiple words, so as to obtain the respective information amount of the two sentences and the total information of the two sentences Quantity; Finally, the similarity of sentences is defined and calculated according to the Jaccard similarity principle. The present invention can realistically simulate human beings' judgment on the similarity of sentences, and does not need to use corpus training parameters or use experience parameters, does not depend on the scale of the corpus, and does not need other natural language processing technologies such as part-of-speech tagging; the time performance is excellent, and it is suitable for general-length sentences. The sentence is right, to obtain quasi-real-time computing efficiency on the current mainstream multi-core PC.

Description

A Calculation Method of Sentence Similarity Based on Information Amount

technical field

The invention relates to a sentence similarity calculation method, in particular to an information-based sentence similarity calculation method, which belongs to the technical field of natural language processing.

Background technique

Sentence or short text similarity calculation is an important research content of natural language processing. In recent years, it has played an increasingly important role in information retrieval, machine translation, question answering system, automatic summarization and other application fields. The traditional method mostly uses the calculation method of document similarity, and only regards the sentence words as meaningless symbols that are not related to each other, which is not accurate enough for the calculation of sentences containing a small number of words. However, the commonly used hybrid methods usually need to train parameters on related data sets or use empirical parameters. The disadvantage is that they rely on training data sets and are not very versatile.

Contents of the invention

The purpose of the method of the present invention is to solve the above-mentioned problem, provide a kind of sentence similarity calculation method based on the amount of information, by using the essential property of the language of the amount of information, use the tolerance-exclusion principle to obtain accurate total information amount of a plurality of words, thereby obtain It is closer to the sentence similarity result obtained by human subjective judgment.

The idea of the inventive method is first to determine the meaning of the word by the concept of the maximum amount of information between two sentence words; then utilize the hierarchical structure of the Semantic Web (such as WordNet) and the statistics of the corpus (such as the BNC corpus or the Brown corpus, etc.) to calculate The amount of information of words and the amount of public information between multiple words; then apply the principle of tolerance and exclusion in combinatorics to calculate the total information amount of multiple words, so as to obtain the respective information amount of two sentences and the total information amount of two sentences ; Finally, define and calculate the sentence similarity according to the Jaccard similarity principle.

In order to achieve the above object, the technical scheme adopted in the present invention is:

A method for calculating sentence similarity based on information content, comprising the following steps:

Step 1: Input the two sentences s _a and s _b to be calculated, and record the sentences s _a and s _b as:

in, and represent the i-th word of sentences s _a and s _b respectively, and n and m represent the number of words of sentences s _a and s _b respectively;

Step 2: Select the meaning of the words in the input sentence, the process is as follows:

words The meaning of the word is determined according to Formula 1:

[Formula 1]

Among them, subsum(c ₁ ,c ₂ ) is the collection of all concepts including concepts c ₁ and c ₂ in the Semantic Web, Represents all words in the Semantic Web that contain The set of concepts, concepts(s _b ) represents the set of concepts in the semantic web that contains all the words in the sentence s _b , P(c) is the frequency of concept c in the corpus, special, if P(c) is 0 , then log P(c) is 0, and the value of P(c) is determined according to formula 2:

[Formula 2]

P(c)＝ _{Σw∈words(c)} count(w)/N

Among them, words(c) represents the set of all words in concept c and all sub-concepts of concept c in the semantic web, count(w) is the frequency of word w in the corpus, and N represents the sum of the frequencies of all concepts in the semantic web, The frequency of each concept is the sum of the total frequencies of all words in the concept in the corpus;

Similarly, in formula 1 replace with concepts(s _b ) is replaced by concepts(s _a ), the meaning of the i-th word in the sentence s _b can be obtained

After the word meaning is determined, the sentences s _a and s _b can be recorded as:

Step 3: According to the sentence obtained in step 2 to determine the meaning of the word, apply the principle of inclusion and exclusion in combinatorics to calculate the respective information content of sentences s _a and s _b and the total information content of the two sentences. The calculation process is as follows:

The calculation formula of the information content IC(s _a ) of the sentence s _a is shown in formula 3:

[Formula 3]

in, Represents the semantic information space jointly constructed by the hierarchical structure of the Semantic Web and corpus statistics, Calculate according to formula 4:

[Formula 4]

in, To include concepts in the Semantic Web The collection of all concepts of

In the same way, replace all letters a in formula 3 and formula 4 with b, and n with m, and the information amount of sentence s _b can be obtained;

The set of all non-repetitive words in sentences s _a and s _b is regarded as a new sentence, and the total information IC(s _a ∪s _b ) of sentences s _a and s _b can be obtained through formula 5:

[Formula 5]

Among them, p is the total number of words that are not repeated in sentences s _a and s _b ;

Step 4: Define the public information amount COMMONIC(s _a , s _b ) of two sentences s _a and s _b from the relationship between union and intersection, and the calculation formula is shown in formula 6:

[Formula 6]

COMMONIC(s _a ,s _b )＝IC(s _a )+IC(s _b )-IC(s _a ∪s _b )

Step 5: According to the Jaccard correlation principle, define the similarity sim(s _a , s _b ) of sentences s _a and s _b , and the calculation formula is shown in formula 7:

[Formula 7]

Step 6: Output the similarity sim(s _a ,s _b ) of the two sentences.

Beneficial effect

Compared with the prior art, the method of the present invention further improves the accuracy of the hybrid method, can realistically simulate human judgment on the similarity of sentences, and does not need to use corpus training parameters or use experience parameters, does not depend on the size of the corpus, and does not require part-of-speech tagging and other natural language processing technologies can be directly used for sentence similarity calculation, which has good versatility; the time performance is excellent, and for sentence pairs of general length, quasi-real-time computing efficiency can be obtained on current mainstream multi-core PCs.

Description of drawings

Figure 1 is a flow chart of the method of the present invention.

Fig. 2 is a comparison diagram between the method of the present invention and other methods.

detailed description

The implementation process of the present invention will be described in detail below in conjunction with the accompanying drawings.

As shown in Figure 1, the inventive method is mainly divided into 5 steps:

Step 1: Enter the two sentences to be calculated. Remember the sentences s _a and s _b are:

in, and represent the i-th word of sentences s _a and s _b respectively, and n and m represent the number of words of sentences s _a and s _b respectively.

Step 2: Due to the common phenomenon of polysemous words, selecting the meaning of the words in the input sentence can eliminate the uncertainty of the sentence semantics, so as to prepare for the subsequent calculation of sentence similarity. The specific process is as follows:

(1) In the two input sentences, select a word to form a word pair;

(2) Use the meaning (or concept) of the word in the semantic network (such as WordNet) to replace the word, and the word pair is converted into multiple word sense pairs;

(3) Calculate the amount of public information between word sense pairs;

(4) choose the maximum amount of public information as the amount of public information of the word pair;

(5) sort the public information of all word pairs in descending order;

(6) define two arrays, be used for recording the meaning of words in each sentence, and all element values in the initialization array are undetermined meanings;

(6) Process according to the order of (5), if the meaning of the word has not been determined, that is, the corresponding element value of the word in the array is still undetermined, then the meaning of the word corresponding to the amount of information is recorded in the meaning array ;

(7) Until the array element values have no undetermined meanings, that is, the meanings of all words are determined.

words The meaning (or concept) of is determined according to formula 1:

[Formula 1]

in

[Formula 2]

P(c)＝ _{∑w∈words(c)} count(w)/N

Here, subsum(c ₁ ,c ₂ ) is the collection of all concepts including concepts c ₁ and c ₂ in the Semantic Web, Represents all contained words in the Semantic Web concepts(s _b ) represents the set of concepts in the semantic web including all the words in sentence s _b , and P(c) is the frequency of concept c in a corpus. In particular, if P(c) is 0, then logP(c) is 0;

words(c) represents the collection of all words in concept c and all sub-concepts of concept c in the semantic web, count(w) is the frequency of word w in a certain corpus, and N represents the sum of the frequencies of all concepts in the semantic web , and the frequency of each concept is the sum of the total frequencies of all words in the concept in the corpus.

In the same way, the meaning of the i-th word in the sentence s _b can be obtained After the word meaning is determined, the sentences s _a and s _b can be recorded as:

Step 3: According to the sentence whose meaning is determined in step 2, apply the principle of inclusion and exclusion in combinatorics to calculate the respective information volumes of sentences s _a and s _b , that is, the calculation formula of the information volume of sentence s _a is shown in formula 3:

[Formula 3]

in

[Formula 4]

here, Represent concepts in the semantic information space jointly constructed by the hierarchical structure of the Semantic Web and corpus statistics the intersection of To include concepts in the Semantic Web The set of all concepts of , for any k between 1 and n, Indicates a combination of k words in the sentence s _a .

In the same way, replace all letters a in formula 3 and formula 4 with b, and n with m, and the information amount of sentence s _b can be obtained.

The set of all non-repetitive words in sentences s _a and s _b is regarded as a new sentence, then the total information amount of sentences s _a and s _b can be obtained through formula 5:

[Formula 5]

Among them, p is the total number of words that are not repeated in the two sentences.

Step 4: The respective information amounts of the two sentences and the total information amount of the two sentences obtained in step 3. The public information COMMONIC(s _a , s _b ) of the two sentences s _a and s _b is obtained from the relationship between the union and the intersection, as shown in Formula 6:

[Formula 6]

COMMONIC(s _a ,s _b )＝IC(s _a )+IC(s _b )-IC(s _a ∪s _b )

Step 5: The total information and public information of the two sentences obtained through steps 3 and 4, according to the Jaccard correlation principle, the similarity between the two sentences s _a and s _b can be calculated by formula 7:

[Formula 7]

As mentioned above, the present invention provides a method for calculating sentence similarity based on information content. Through the real sentence pair input by the user, the system will automatically calculate the result that is realistic to the human judgment of sentence similarity.

The comparison results between the method of the present invention and other four existing sentence similarity calculation methods are given below. The semantic web WordNet and corpus BNC are used in the experiment. Pearson correlation coefficient (PCC) for linear correlation, Spearman rank correlation coefficient (SRCC) for general rank correlation and Kendall rank correlation coefficient (KRCC) for probability rank correlation were used for evaluation. Table 1 shows the scoring results of the sentence pairs corresponding to the words using various methods.

Table 1 Manual scoring of sentence pairs and scoring by various methods

Figure 2 can be obtained from Table 1. As can be seen from Figure 2, the method in this paper is superior to the other four methods on PCC, SRCC and KRCC, indicating that the model's judgment of sentence similarity is closer to human subjective judgment. In addition, the average value of PCC between the scores of a single volunteer and the mean of all volunteers in the artificial data set is 0.825, and the maximum value is 0.921; while the PCC of the method of the present invention is higher than the average value of single-person scoring and lower than the average value of single-person scoring The upper limit value of ; it shows that the judgment level of the model on sentence similarity is higher than the average human level, and the result is credible.

The specific description above further elaborates the purpose, technical solution and beneficial effect of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not used to limit the protection of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims (1)

1. A method for calculating sentence similarity based on information volume, characterized in that, comprising the following steps: Step 1: Input the two sentences s _a and s _b to be calculated, and record the sentences s _a and s _b as: ${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \}</span>$ ${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \}</span>$ in, and represent the i-th word of sentences s _a and s _b respectively, and n and m represent the number of words of sentences s _a and s _b respectively; Step 2: Select the meaning of the words in the input sentence, the process is as follows: words The meaning of the word is determined according to Formula 1: [Formula 1] ${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; arg</span>}\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>$ c∈subsum(c ₁ ,c ₂ ) ${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; )</span>$ c ₂ ∈ concepts(s _b ) Among them, subsum(c ₁ ,c ₂ ) is the collection of all concepts including concepts c ₁ and c ₂ in the Semantic Web, Represents all words in the Semantic Web that contain The set of concepts, concepts(s _b ) represents the set of concepts in the semantic web that contains all the words in the sentence s _b , P(c) is the frequency of concept c in the corpus, special, if P(c) is 0 , then logP(c) is 0, and all values of P(c) are determined according to formula 2: [Formula 2] P(c)＝ _{∑w∈words(c)} count(w)/N Among them, words(c) represents the collection of all words in concept c and all sub-concepts of concept c in the semantic web, count(w) is the frequency of word w in the corpus, and N represents the sum of the frequencies of all concepts in the semantic web, The frequency of each concept is the sum of the total frequencies of all words in the concept in the corpus; Similarly, in formula 1 replace with concepts(s _b ) is replaced by concepts(s _a ), the meaning of the i-th word in the sentence s _b can be obtained After the word meaning is determined, the sentences s _a and s _b can be recorded as: ${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \}</span>$ ${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \}</span>$ Step 3: According to the sentence obtained in step 2 to determine the meaning of the word, apply the principle of inclusion and exclusion in combinatorics to calculate the respective information content of sentences s _a and s _b and the total information content of the two sentences. The calculation process is as follows: The calculation formula of the information content IC(s _a ) of the sentence s _a is shown in formula 3: [Formula 3] $\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; )</span>$ in, Represents the semantic information space jointly constructed by the hierarchical structure of the Semantic Web and corpus statistics, Calculate according to formula 4: [Formula 4] $\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; a</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">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; b</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \&lsqb;</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">\; c</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span>$ in, To include concepts in the Semantic Web The collection of all concepts of In the same way, replace all letters a in formula 3 and formula 4 with b, and n with m, and the information amount of sentence s _b can be obtained; Considering the set of all non-repetitive words in sentences s _a and s _b as a new sentence, the total information IC(s _a ∪s _b ) of sentences s _a and s _b can be obtained through formula 5: [Formula 5] $\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; \&cup;</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; )</span>$ Among them, p is the total number of words that are not repeated in sentences s _a and s _b ; Step 4: Define the public information amount COMMONIC(s _a , s _b ) of two sentences s _a and s _b from the relationship between union and intersection, and the calculation formula is shown in formula 6: [Formula 6] COMMONIC(s _a ,s _b )＝IC(s _a )+IC(s _b )-IC(s _a ∪s _b ) Step 5: According to the Jaccard correlation principle, define the similarity sim(s _a , s _b ) of sentences s _a and s _b , and the calculation formula is shown in formula 7: [Formula 7] $\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; C</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; N</span>}\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; I</span>}\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; \&cup;</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; )</span>}$ Step 6: Output the similarity sim(s _a ,s _b ) of the two sentences.