CN108629019B - Question-answer field-oriented question sentence similarity calculation method containing names - Google Patents
Question-answer field-oriented question sentence similarity calculation method containing names Download PDFInfo
- Publication number
- CN108629019B CN108629019B CN201810433143.XA CN201810433143A CN108629019B CN 108629019 B CN108629019 B CN 108629019B CN 201810433143 A CN201810433143 A CN 201810433143A CN 108629019 B CN108629019 B CN 108629019B
- Authority
- CN
- China
- Prior art keywords
- question
- similarity
- sentence
- corpus
- sim
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004364 calculation method Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 24
- 230000002194 synthesizing effect Effects 0.000 claims description 7
- 230000001105 regulatory effect Effects 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 8
- 238000007796 conventional method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 240000000146 Agaricus augustus Species 0.000 description 1
- 240000005373 Panax quinquefolius Species 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a question similarity calculation method containing names in the question-answering field, which respectively calculates the similarity between the names and the non-names, calculates the sentence similarity from the two aspects of the word order and the length of the sentence by considering the sentence structure, and finally obtains the similarity of the whole sentence according to the sentence semantic similarity and the structural similarity in a weighting manner. The problem that the sentences containing names can not be judged according with human subjectivity when the sentences are calculated by the conventional sentence similarity calculation method is solved. The method provided by the invention can more accurately calculate the similarity of sentences containing names of people and can be widely applied to the field of question answering.
Description
Technical Field
The invention relates to the technical field of question answering, in particular to a question sentence similarity calculation method containing names in the field of question answering.
Background
The similarity calculation of question sentences is always the basic and important research work in the fields of artificial intelligence and natural language processing, is also a research hotspot, and has very wide application, such as question-answering systems, information retrieval systems and the like.
The algorithms related to the similarity calculation of Chinese sentences at present can be roughly divided into the following categories: the first category is feature word based methods; the second category is sentence structure based methods; the third category is semantic dictionary based methods. Firstly, the first method is based on a method of characteristic words, which is to extract the characteristic words of two question sentences to be compared respectively, then compare the characteristic words, calculate the similarity of the characteristic words, and express the similarity of the two question sentences by using the similarity result. Next to the second method, a method based on a syntactic structure, which refers to calculating the similarity of the syntactic structures of two sentences by analyzing the structures of the two sentences. By comparing the part of speech sequences of the two sentences, after the optimal identical part of speech sequence is matched, the similarity between the part of speech and the part of speech is compared, so that the similarity between the two sentences is reflected. The third method, a method based on semantic dictionary, refers to reflecting the similarity between two sentences through the similarity of words in question sentences. When the word similarity is calculated, a large-scale semantic dictionary is used, for example, a How Net knowledge base is used for calculating the similarity of two question sentences, the similarity is calculated by matching all the words in the two question sentences pairwise, the two words with the highest similarity result are used as an optimal matching pair, and finally the result of weighted average of the similarity of all the optimal matching pairs of words represents the semantic similarity of the whole sentence.
However, when the similarity of sentences containing names of people is calculated, the above three methods cannot accurately calculate the questions containing names of people, for example, for two questions, "zhrenchang is the fourth generation royal grandpa of jing jiang dynasty" and "zhrencheng is the fourth generation royal grandpa of jing jiang dynasty", if the three methods introduced before are used for calculation, the similarity of the two sentences obtained is extremely high. However, from the actual angle, zhren chang and zhren sanden are two individuals, which are the princes of jing fu and have similar names but different actual meanings. Therefore, the similarity result calculated by the method is extremely high and does not accord with the subjective judgment of people.
Disclosure of Invention
The invention aims to solve the problem that the difference between names and the importance of the names to the whole sentence cannot be reflected when the similarity of the sentences is calculated in the current question-answering field, so that the result of the question-answering similarity calculation is poor and glad.
In order to solve the problems, the invention is realized by the following technical scheme:
a question sentence similarity calculation method containing names in the question and answer field specifically comprises the following steps:
step 1, calculating current input question L and each question S in corpuszThe sentence structure similarity specifically includes:
step 1.1, calculating sentence length similarity SimLen(L,Sz)：
Wherein LenLThe number of the words after the input question sentence L is segmented is shown,
representing a corpus question SzThe number of words after word segmentation;
step 1.2, calculating sentence word order similarity SimOrd(L,Sz)：
Where RevOrd indicates that the same term is in corpus question S relative to input question LzIn (3), MaxRevOrd indicates that the same word number sequence is in the corpus question S relative to the input question LzThe maximum inverse number of;
step 1.3, synthesizing sentence length similarity Sim obtained in step 1.1Len(L,Sz) And the sentence word order similarity Sim obtained in step 1.2Ord(L,Sz) To obtain the current input question L and each question S in the corpuszSentence structure similarity Simstru(L,Sz)；
Simstru(L,Sz)＝μ1SimLen(L,Sz)+μ2SimOrd(L,Sz)
Wherein, mu1Weight, μ, representing sentence length similarity2Weight, μ, representing sentence structural similarity1+μ2＝1；
Step 2, calculating the current input question L and each question S in the corpuszThe sentence semantic similarity specifically includes:
step 2.1, calculating the similarity of names and words of sentences
Wherein x is1And y1Respectively representing the year and month of birth of the name in the input question L, x2And y2Respectively representing corpus question SzYear and month of birth of the Chinese name, p1And q is1Respectively representing the year and month of birth of the name spouse in the input question L, p2And q is2Respectively representing corpus question SzThe birth year and the birth month of the middle-name spouse, alpha is the regulating parameter of the human, beta is the regulating parameter of the human spouse, and alpha + beta is 1;
step 2.2, calculating similarity of non-name words of sentences
Wherein, C1iIndicating the words L in the input question Lv1A certain meaning item of (1), C2jRepresenting a corpus question SzChinese word
N represents the word L in the input question Lv1The number of semantic items, m, representing a corpus question SzChinese word
Number of middle terms, Sim (C)1i,C2j) Representing an item of significance C1iAnd item of sense C2jThe similarity of (2);
step 2.3, synthesizing the similarity of names and words of the sentences obtained in the step 2.1
And the similarity of the non-name words of the sentences obtained in the step 2.2
Obtaining the current input question L and each question S in the corpuszSemantic similarity Sim of sentencessem(L,Sz)；
Wherein a represents an input question L and a corpus question SzB represents the logarithm of the best matching pair obtained in the non-human name set of input question L and corpus question, γ1Weight, gamma, representing the similarity of the names and words2Weight, gamma, representing similarity of non-human name words1+γ2＝1；
Step 3, synthesizing sentence structure similarity Sim obtained in step 1stru(L,Sz) And step 2, obtaining the semantic similarity Sim of the sentencessem(L,Sz) To obtain the current input question L and each question S in the corpuszGlobal sentence similarity Sim (L, S)z)：
Sim(L,Sz)＝λ1Simstru(L,Sz)+λ2Simsem(L,Sz)
Wherein λ is1Representing sentence structure phaseWeight of similarity, λ2Weight, λ, representing semantic similarity of sentences1+λ2＝1；
Step 4, the whole sentence similarity Sim (L, S) obtained in the step 3z) Sorting and selecting the whole sentence similarity Sim (L, S) with the current input question L from the corpusz) Outputting the highest question as a question similarity calculation result;
s abovezRepresents the Z-th sentence in the corpus, Z belongs to (1,2, …, Z), and Z is the number of question sentences in the corpus.
In the step 2.2, the similarity of the non-famous words of the sentence can be calculated by utilizing different Guinea electricity knowledge bases
But preferably calculates similarity of non-name words of sentences by using the How Net knowledge base
Firstly, when the similarity of question sentences containing names is calculated, the names and the names of people are distinguished, and the similarity of the names and the names of people is calculated respectively; then, considering the structure of the sentence, and calculating the similarity of the sentence from the two aspects of the word order and the length of the sentence; and finally, weighting according to the semantic similarity and the structural similarity of the sentence to obtain the similarity of the whole sentence. The invention solves the problem that the prior sentence similarity calculation method can not obtain the subjective judgment conforming to the human name when calculating the sentences containing the names.
Compared with the prior art, the invention provides a method for dividing the sentence into a part with the name of a person and a part with the name of a non-person and respectively calculating the similarity, simultaneously considers the semantics of the sentence and the structural similarity of the sentence, solves the problems that the calculated similarity is not accurate or does not accord with the subjective judgment of a person when the sentence with the name of the person is involved in the prior art, and has good practicability.
Drawings
Fig. 1 is a partial example diagram of an example sentence divided into a person name and a non-person name.
Fig. 2 is a flow chart of a method for calculating similarity of question sentences according to names of people in the invention.
Fig. 3a is an exemplary diagram of calculating question similarity according to a conventional method based on feature words.
Fig. 3b is a diagram illustrating a method for calculating question similarity according to a conventional syntax-based structure.
Fig. 3c is a schematic diagram of calculating question similarity according to a conventional method based on a semantic dictionary.
Fig. 4 is a schematic diagram of calculating the similarity of question sentences containing names of people according to the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings in conjunction with specific examples.
In the existing question similarity calculation technology, the special importance of the names of the persons is not considered when the question similarity containing the names of the persons is calculated, and if the names of the persons are not processed properly, the difference between the names of the persons and the importance of the names of the persons to the whole sentence cannot be reflected, so that the result of the question similarity calculation is poor. The invention fully considers the important role of the names of people in calculating the similarity of the question, and respectively calculates the similarity according to the fact that the question is divided into two parts of the names of people and the names of non-people. Meanwhile, the influence of the sentence structure on the sentence similarity is considered. Fig. 1 is an example diagram in which example sentences "zhu cheng is the first generation royal of jing jiang dynasty" and "zhu chang is the first generation royal of jing jiang dynasty" are divided into two parts, and similarity is calculated. Wherein the circles represent "sanden" and "sandong chang", represent the named part of the sentence, and the boxes represent the non-named part of the sentence.
The invention firstly considers the question-answering field, and relates to calculating the similarity between related name question sentences by obtaining the birth year and month of the character to calculate the similarity between name words, so as to avoid the same birth year and month of the character, thereby increasing the birth year and month of the spouse for supplement; when the semantic similarity of sentences is calculated, the importance of names to the calculation of the similarity of the sentences is embodied by giving different weights to the names and the non-names; and the factors of sentence structure are considered, and the calculation is respectively carried out according to the length and the word sequence of the sentence; and finally, giving different weights to the similarity of sentence semantics and structure to obtain the overall similarity of the sentences.
A question sentence similarity calculation method containing names in the question and answer field is shown in a flow diagram of fig. 2, and comprises the following steps:
step 1, calculating current input question L and each question S in corpuszThe sentence structure similarity. For the sentence structure similarity calculation, the similarity calculation is mainly performed on the sentence structure from two aspects, namely the length of the sentence and the word order of the sentence.
Step 1.1, calculating sentence length similarity SimLen(L,Sz)：
Wherein LenLThe number of the words after the word segmentation of the current input question sentence L is shown,
representing question S in corpuszThe number of the words after word segmentation, wherein SzRepresents the Z-th sentence in the corpus, Z belongs to (1,2, …, Z), and Z is the number of question sentences in the corpus.
Step 1.2, calculating sentence word order similarity SimOrd(L,Sz)：
In the formula, assuming that the sequence of the same words of two sentences is positive in the input question L, RevOrd represents the question S of the same words in the corpuszIn (1)The inverse number, MaxRevOrd, represents the maximum inverse number of the same word number sequence;
step 1.3, the similarity between the sentence structure and the word order is integrated, the structural similarity of the whole sentence can be calculated, and the calculation formula is as follows:
Simstru(L,Sz)＝μ1SimLen(L,Sz)+μ2SimOrd(L,Sz)
wherein, mu1Weight, μ, representing sentence length similarity2Weight representing sentence structural similarity, and μ1+μ2＝1。
Step 2, calculating the current input question L and each question S in the corpuszSemantic similarity of sentences. Similarity calculation is carried out aiming at the semantics of sentences, and the similarity calculation mainly comprises two parts, namely human name similarity calculation and non-human name similarity calculation.
Step 2.1, calculating the similarity of the names of the people;
step 2.1.1, using the birth year and month of the person as a vector coordinate, and adopting a calculation formula as follows:
the birth year and the birth month of the person are used as the name Lx of the person in the input question L1And question S in corpuszName of Chinese
The vector coordinates of (1), wherein x is year coordinate, y is month coordinate, and cosine value is used for representing the similarity of two names in the birth year and month;
step 2.1.2, considering the situation that two people can be born in the same year and month, the birth year and month of the spouse is added for supplement, and the calculation formula is as follows:
match figures with a dollThe year and month of birth of (1) as the name of the person in the input question Lu1And question S in corpuszName of Chinese
Wherein p is a year coordinate, q is a month coordinate, and the similarity of the birth year and month of the two spouses is expressed by using cosine values;
step 2.1.3, integrating the two factors which have important influence on the similarity of the names to obtain the overall similarity of the names, wherein the calculation formula is as follows:
weighting and summing the year and month of birth similarity of the characters and the character spouses to obtain a value which is the overall similarity of the names of the characters, wherein alpha and beta are adjusting parameters, and alpha + beta is 1;
step 2.2, calculating the similarity of the non-name words;
the method for calculating the similarity of words by using the How Net has the following calculation formula
Wherein L isv1And
respectively an input question L and a question S in the corpuszTwo words of (A), C1iFor inputting question L words Lv1A certain meaning item of (1), C2jAs question S in corpuszWord and phrase
The maximum similarity between all the meaning items of the two terms is used for representing the similarity of the non-human name terms;
step 2.3, the calculation results of the similarity of the names and the similarity of the non-names are integrated, the semantic similarity of the whole sentence can be calculated, and the calculation formula is as follows:
wherein a represents an input question L and a question S in the corpuszTo obtain the logarithm of the best matching pair, Lu1Representing the name of the sentence L in the pair of names,
representing sentences S in paired nameszB represents the logarithm of the best matching pair obtained from the input sentence and the non-human name set of sentences in the corpus, Lv1A word representing the sentence L in the pair of words,
in sentence S for paired wordszTerm of γ1Representing the weight given to the term of a person, gamma2Denotes a weight given to a non-human term, and γ1+γ2＝1。
Step 3, synthesizing sentence structure similarity Sim obtained in step 1stru(L,Sz) And step 2, obtaining the semantic similarity Sim of the sentencessem(L,Sz) To obtain the current input question L and each question S in the corpuszGlobal sentence similarity Sim (L, S)z)。
And (3) calculating the similarity of the whole sentence, namely integrally calculating the similarity of the names of the people and the similarity of the names of the non-people, wherein the calculation formula is as follows:
Sim(L,Sz)＝λ1Simstru(L,Sz)+λ2Simsem(L,Sz)
wherein λ is1Weight, λ, representing sentence structure similarity2Weight representing semantic similarity of sentences, and1+λ2＝1。
and 4, sequencing the overall sentence similarity obtained in the step 3, and selecting the question with the highest overall sentence similarity with the current input question L from the corpus, and outputting the question as a question similarity calculation result.
When the method is adopted to calculate the sentence similarity, the condition of sentences containing names is considered, and the similarity of the semantics and the structure of the sentences is considered, so that the sentence similarity calculated by using the model is more accurate and effective.
Fig. 3 is a diagram illustrating the calculation of sentence similarity according to three conventional methods. In the method based on the feature words in fig. 3a, only the occurrence frequencies of words such as "zhuochang", "zhuocheng", "jing jiang fu", etc. in sentences are considered when calculating the sentence similarity, and the semantics cannot be well processed; fig. 3b mainly considers the structural features of sentences when calculating the sentence similarity by the syntax structure-based method, and does not analyze semantic information well, so that for two sentences "zhu chang is the royal of the first generation of jing jiang coworker" and "zhu chang is the royal of the second generation of jing jiang coworker", the structures of the sentences are the same, and then the similarity of the two question sentences is considered to be the same; although the semantic dictionary-based method in fig. 3c can effectively understand semantic information, for the large-scale knowledge base How Net does not completely cover all the names of people, nor does it clearly distinguish similar names, so for two names of "zhu-anchang" and "zhu-ren sheng", if there are two names in the knowledge base, the similarity is extremely high, and if there are no two names, the similarity is extremely low. In fact, "Zhu ren Chang" and "Zhu ren Cheng" are two completely different people.
In fig. 4, the situation that names are included in sentences is considered, when names are included in question sentences, the similarity between two names of "zhu-anchang" and "zhu-ancheng" is calculated separately, the similarity between the remaining half sentences is calculated at the same time, and after semantic analysis, sentence structure analysis is performed to calculate the word order and length similarity of the sentences.
The invention provides a question similarity calculation method containing names of people, which comprises the following steps: and respectively calculating the similarity of the human name and the non-human name, calculating the similarity of sentences from the two aspects of the word order and the length of the sentences by considering the structure of the sentences, and finally weighting according to the semantic similarity and the structural similarity of the sentences to obtain the similarity of the whole sentences. The problem that the sentences containing names can not be judged according with human subjectivity when the sentences are calculated by the conventional sentence similarity calculation method is solved. The method provided by the invention can more accurately calculate the similarity of sentences containing names of people and can be widely applied to the field of question answering.
It should be noted that, although the above-mentioned embodiments of the present invention are illustrative, the present invention is not limited thereto, and thus the present invention is not limited to the above-mentioned embodiments. Other embodiments, which can be made by those skilled in the art in light of the teachings of the present invention, are considered to be within the scope of the present invention without departing from its principles.
Claims (2)
1. A question sentence similarity calculation method containing names in the question and answer field is characterized by comprising the following steps:
step 1, calculating current input question L and each question S in corpuszThe sentence structure similarity specifically includes:
step 1.1, calculating sentence length similarity SimLen(L,Sz)：
Wherein LenLThe number of the words after the input question sentence L is segmented is shown,
representing a corpus question SzThe number of words after word segmentation;
step 1.2, calculating sentence word order similarity SimOrd(L,Sz)：
Where RevOrd indicates that the same term is in corpus question S relative to input question LzIn (3), MaxRevOrd indicates that the same word number sequence is in the corpus question S relative to the input question LzThe maximum inverse number of;
step 1.3, synthesizing sentence length similarity Sim obtained in step 1.1Len(L,Sz) And the sentence word order similarity Sim obtained in step 1.2Ord(L,Sz) To obtain the current input question L and each question S in the corpuszSentence structure similarity Simstru(L,Sz)；
Simstru(L,Sz)＝μ1SimLen(L,Sz)+μ2SimOrd(L,Sz)
Wherein, mu1Weight, μ, representing sentence length similarity2Weight, μ, representing sentence structural similarity1+μ2＝1；
Step 2, calculating the current input question L and each question S in the corpuszThe sentence semantic similarity specifically includes:
step 2.1, calculating the similarity of names and words of sentences
Wherein x is1And y1Respectively representing the year and month of birth of the name in the input question L, x2And y2Respectively representing corpus question SzYear and month of birth of the Chinese name, p1And q is1Respectively representing the year and month of birth of the name spouse in the input question L, p2And q is2Respectively representing corpus question SzThe birth year and the birth month of the middle-name spouse, alpha is the regulating parameter of the humanNumber, β is a regulatory parameter of the human spouse, α + β ═ 1;
step 2.2, calculating similarity of non-name words of sentences
Wherein, C1iIndicating the words L in the input question Lv1A certain meaning item of (1), C2jRepresenting a corpus question SzChinese word
N represents the word L in the input question Lv1The number of semantic items, m, representing a corpus question SzChinese word
Number of middle terms, Sim (C)1i,C2j) Representing an item of significance C1iAnd item of sense C2jThe similarity of (2);
step 2.3, synthesizing the similarity of names and words of the sentences obtained in the step 2.1
And the similarity of the non-name words of the sentences obtained in the step 2.2
Obtaining the current input question L and each question S in the corpuszSemantic similarity Sim of sentencessem(L,Sz)；
Wherein a represents an input question L and a corpus question SzB represents the logarithm of the best matching pair obtained in the non-human name set of input question L and corpus question, γ1Weight, gamma, representing the similarity of the names and words2Weight, gamma, representing similarity of non-human name words1+γ2＝1；
Step 3, synthesizing sentence structure similarity Sim obtained in step 1stru(L,Sz) And step 2, obtaining the semantic similarity Sim of the sentencessem(L,Sz) To obtain the current input question L and each question S in the corpuszGlobal sentence similarity Sim (L, S)z)：
Sim(L,Sz)＝λ1Simstru(L,Sz)+λ2Simsem(L,Sz)
Wherein λ is1Weight, λ, representing sentence structure similarity2Weight, λ, representing semantic similarity of sentences1+λ2＝1；
Step 4, the whole sentence similarity Sim (L, S) obtained in the step 3z) Sorting and selecting the whole sentence similarity Sim (L, S) with the current input question L from the corpusz) Outputting the highest question as a question similarity calculation result;
s abovezRepresents the Z-th sentence in the corpus, Z belongs to (1,2, …, Z), and Z is the number of question sentences in the corpus.
2. The method for calculating question similarity containing names in question-answering field according to claim 1, wherein in step 2.2, the How-similar non-name words in sentences are calculated by using How Net knowledge base
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810433143.XA CN108629019B (en) | 2018-05-08 | 2018-05-08 | Question-answer field-oriented question sentence similarity calculation method containing names |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810433143.XA CN108629019B (en) | 2018-05-08 | 2018-05-08 | Question-answer field-oriented question sentence similarity calculation method containing names |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108629019A CN108629019A (en) | 2018-10-09 |
| CN108629019B true CN108629019B (en) | 2021-04-30 |
Family
ID=63695949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810433143.XA Active CN108629019B (en) | 2018-05-08 | 2018-05-08 | Question-answer field-oriented question sentence similarity calculation method containing names |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108629019B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472008A (en) * | 2018-11-20 | 2019-03-15 | 武汉斗鱼网络科技有限公司 | A kind of Text similarity computing method, apparatus and electronic equipment |
| CN111221954A (en) * | 2020-01-09 | 2020-06-02 | 珠海格力电器股份有限公司 | Method, device, storage medium and terminal for constructing household appliance maintenance question-answer library |
| CN111666770B (en) * | 2020-06-02 | 2023-07-18 | 泰康保险集团股份有限公司 | Semantic matching method and device |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9424298B2 (en) * | 2014-10-07 | 2016-08-23 | International Business Machines Corporation | Preserving conceptual distance within unstructured documents |
| CN106649404B (en) * | 2015-11-04 | 2019-12-27 | 陈包容 | Method and device for creating session scene database |
| US20170212872A1 (en) * | 2016-01-22 | 2017-07-27 | International Business Machines Corporation | Duplicate post handling with natural language processing |
| CN107729392B (en) * | 2017-09-19 | 2020-07-10 | 广州市妇女儿童医疗中心 | Text structuring method, device and system and non-volatile storage medium |
-
2018
- 2018-05-08 CN CN201810433143.XA patent/CN108629019B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108629019A (en) | 2018-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10860641B2 (en) | Method, apparatus, and electronic devices for searching images | |
| CN109299480B (en) | Context-based term translation method and device | |
| WO2017162134A1 (en) | Electronic device and method for text processing | |
| CN110675269B (en) | Text auditing method and device | |
| CN104050302B (en) | Topic detecting system based on atlas model | |
| CN106610951A (en) | Improved text similarity solving algorithm based on semantic analysis | |
| CN108629019B (en) | Question-answer field-oriented question sentence similarity calculation method containing names | |
| CN105760474A (en) | Document collection feature word extracting method and system based on position information | |
| CN111626042B (en) | Reference digestion method and device | |
| CN108388554A (en) | Text emotion identifying system based on collaborative filtering attention mechanism | |
| CN105069647A (en) | Improved method for extracting evaluation object in Chinese commodity review | |
| Al-Azzawy et al. | Arabic words clustering by using K-means algorithm | |
| CN113076744A (en) | Cultural relic knowledge relation extraction method based on convolutional neural network | |
| CN110866087B (en) | An entity-oriented text sentiment analysis method based on topic model | |
| CN110309513B (en) | Text dependency analysis method and device | |
| Sujana et al. | LiDA: Language-independent data augmentation for text classification | |
| Sitorus et al. | Sensing trending topics in twitter for greater Jakarta area | |
| CN113761104A (en) | Method and device for detecting entity relationship in knowledge graph and electronic equipment | |
| CN112287667A (en) | Text generation method and equipment | |
| Aktas et al. | Text classification via network topology: A case study on the holy quran | |
| CN107818078B (en) | Semantic association and matching method for Chinese natural language dialogue | |
| CN104965818A (en) | Project name entity identification method and system based on self-learning rules | |
| CN110674871B (en) | Translation-oriented automatic scoring method and automatic scoring system | |
| Liu et al. | Text-Segment Interaction for Authorship Verification using BERT-based Classification. | |
| Mishina et al. | Word sense disambiguation of adjectives using dependency structure and degree of association between sentences |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181009 Assignee: Guilin Biqi Information Technology Co.,Ltd. Assignor: GUILIN University OF ELECTRONIC TECHNOLOGY Contract record no.: X2023980045831 Denomination of invention: A Similarity Calculation Method for Question Sentences with Person Names in the Question Answering Domain Granted publication date: 20210430 License type: Common License Record date: 20231107 |
|
| EE01 | Entry into force of recordation of patent licensing contract |