CN111291156B - Knowledge graph-based question and answer intention recognition method - Google Patents
Knowledge graph-based question and answer intention recognition method Download PDFInfo
- Publication number
- CN111291156B CN111291156B CN202010068158.8A CN202010068158A CN111291156B CN 111291156 B CN111291156 B CN 111291156B CN 202010068158 A CN202010068158 A CN 202010068158A CN 111291156 B CN111291156 B CN 111291156B
- Authority
- CN
- China
- Prior art keywords
- topic
- word
- text
- relationship
- words
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 239000013598 vector Substances 0.000 claims abstract description 17
- 238000004458 analytical method Methods 0.000 claims abstract description 14
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 2
- 238000004220 aggregation Methods 0.000 claims description 2
- 238000011176 pooling Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims description 2
- 238000005065 mining Methods 0.000 claims 1
- 108020001568 subdomains Proteins 0.000 claims 1
- 230000006399 behavior Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3335—Syntactic pre-processing, e.g. stopword elimination, stemming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/374—Thesaurus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a knowledge graph-based question and answer intention recognition method, which comprises the following steps: constructing a domain topic dictionary; constructing a template; performing part-of-speech analysis and syntactic dependency analysis on the search statement; calculating the similarity between the search text and the template sample by using a word vector and an LDA algorithm; based on the known word vector, the TextCNN is used for more extensive intention recognition as a result of open information; and extracting keywords from the domain map. The invention integrates various methods to realize intention recognition, combines the knowledge graph of the field, achieves comprehensive retrieval of accurate answers and related information, and meets various requirements of users.
Description
Technical Field
The invention relates to the technical field of natural language processing and deep learning, in particular to a knowledge-graph-based question and answer intention recognition method.
Background
With the explosion of modern information and the rapid development of science and technology, the functions of a search engine are no longer satisfied with the traditional mode that a large number of related keyword webpages or links are returned after the keywords of the search engine are input, and in this mode, users also need to manually distinguish and browse in the returned results, and redundant results often need to waste a large amount of time of the users to exclude, so that the accuracy is not high. The easier-to-use search engine allows the user to obtain answers or related knowledge with higher directivity after inputting the spoken descriptive information, which requires the program to analyze the user's intention more accurately and efficiently. So more and more intelligent question-answering systems have been put into development and use in recent years.
The traditional intention recognition mainly applies a method of grammar semantic analysis, grammar dependency analysis and template matching, and the method needs a large amount of rules to support, has higher recognition accuracy in a specified range, has poor expansibility, needs to be reconfigured once the rules are changed, and consumes a large amount of manpower.
After machine learning is developed on a large scale, algorithms for classifying problems, such as SVM, bayesian algorithm, KNN, etc., are also widely applied to the field of intention recognition, but the method still faces the problem that manual classification labeling requires a lot of manpower, and in many application scenarios, text input by a user is not a single intention, and one sentence may contain two or more intentions. With the intensive study of machine learning, many students at home and abroad propose a multi-label classification method for solving the problems.
After the deep learning concept is introduced, deep learning algorithms such as CNN, RNN, LSTM, attention mechanism and the like are also widely applied to the text classification field, and many field recognition systems apply the algorithms to practical application in a dispute manner, and make corresponding adjustment and improvement in combination with practical requirements.
By combining the characteristics of the traffic field, the method integrates various methods to realize intention recognition, and combines the knowledge graph of the field to realize comprehensive retrieval of accurate answers and related information, thereby meeting various demands of users.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a knowledge graph-based question and answer intention recognition method.
The aim of the invention is achieved by the following technical scheme:
a knowledge graph-based question and answer intention recognition method comprises the following steps:
a, constructing a domain topic dictionary;
b, constructing a template;
c, performing part-of-speech analysis and syntactic dependency analysis on the search statement;
d, calculating the similarity between the search text and the template sample by using a word vector and an LDA algorithm;
e, based on the known word vector, performing broader intention recognition based on the textCNN to serve as a result of the open information;
f, extracting keywords from the domain map.
One or more embodiments of the present invention may have the following advantages over the prior art:
and integrating various methods to realize intention recognition, combining knowledge patterns of the field, achieving comprehensive retrieval of accurate answers and related information, and meeting various requirements of users.
Drawings
Fig. 1 is a flowchart of a knowledge-graph-based question and answer intention recognition method.
Fig. 2, 3 and 4 are illustrations of specific embodiments.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the following examples and the accompanying drawings.
As shown in fig. 1, the flow of the knowledge-graph-based question and answer intention recognition method comprises the following steps:
step 10, constructing a domain topic dictionary;
because of the many specialized vocabularies, entity names and short names in different specialized fields, the topic dictionary assistance in related vocabulary construction fields needs to be arranged in advance for intention recognition. In many researches about intention recognition, the construction of a topic dictionary in the professional field is lacking, and the lack of flexibility in the manner is considered to require manual maintenance, but in an application scene that a user needs to accurately and quickly search within a specified range, the topic dictionary in the professional field is one of effective means for helping to improve the accuracy and recall rate.
Step 20, constructing a template;
at present, the common viewpoint is to divide the input of a question-answering system into deterministic information under the information category, such as information atmosphere information category, navigation category, transaction category and the like contained in the search behavior of a user, so that the user hopes to obtain a definite answer through a certain question; open information, i.e. information that a user needs to obtain in a certain direction widely or more deeply.
Templates primarily assist the system in resolving deterministic information intent recognition, which may include multiple specific entries associated with intent in existing data, if not differentiated, the results of the search remain multi-directional and may differ. To help discern such intent, a portion of the ambiguous results need to be specified to help accurately identify the user intent.
Step 30, performing part-of-speech analysis and syntactic dependency analysis on the search statement;
whether for Chinese or foreign language retrieval, certain grammatical rules are followed between the words of the input sentence, and understanding these rules helps the program understand the relationships between the words.
Limited to the length of the search term, the general search text is a short term containing a large number of key words. Several types of syntactic relations of major concern include mainly master-predicate relations (SBV), guest-moving relations (VOB), centering relations (ATT), in-state relations (ADV), parallel relations (COO), etc. The roles played by the words with different parts of speech in the sentences are generally fixed, the parts of speech of the words can be known to further divide the components of the sentences on the basis of syntactic dependency analysis, the relation among the words is known, and possible intentions and limitation conditions for the intentions are acquired.
In the traditional method, parts of speech or syntactic dependency is used independently, but in practical application, certain verbs or adverbs can become keywords, key information can be omitted simply by means of identifying entities or dividing sentence components, and in addition, the dividing accuracy of the traditional syntactic dependency analyzer on a plurality of sentence patterns cannot meet the requirement, so that the relation of the words and the parts of speech is required to be analyzed specifically on the basis of acquiring the relation and the parts of speech, and more topics are matched better by combining a constructed domain topic dictionary.
If (part of speech of word is time, place name, person name, etc.):
filling corresponding slots
If (there is a relationship between the main and guest)
If (subject, object and subject dictionary match)
Judging specific retrieval operation according to predicates, and generating a conditional statement together with the object of the subject
Else
Open matching as keywords into the map
Elif (there is a centering relationship, a middle relationship)
If (fixed language, scholarly language and theme dictionary match)
Performing search operation by using the fixed language and the scholarly language as limiting conditions
Elif (fixed language, vernix is a possible aggregation operation)
Conditional definition of keywords for fixed language and scholarly modified language
Else
Open matching as keywords into the map
Through the steps, the program can acquire more possible question-answer accurate search intentions of the user, and does not simply perform combined search matching on keywords existing in sentences.
Step 40, calculating the similarity between the search text and the template sample by using a word vector and an LDA algorithm;
and calculating the similarity between the search text and the template sample by using the word vector and the LDA algorithm.
The professional data and the general data in the application field are used for generating word vectors through word segmentation and text of work such as stop words. Converting words into dense vectors and judging the similarity of words according to the distance between word vectors, but in many cases, two sentences have no obvious relation on the constituent words, so that implicit dirichlet allocation (LDA) needs to be introduced to help a program better understand possible relations between sentences.
LDA is based on Bayesian model, and adopts the premise of prior probability, and the probability is gradually updated according to the obtained posterior distribution in the training process. The first page in the LDA model also needs to preset a possible number of topics, after the assumed number of topics is provided, the prior distribution of the assumed texts accords with the Dirichlet distribution, and the method has the advantages that for any text
θm＝Dirichlet(α → )
Let the words in the text also conform to the Dirichlet distribution, i.e. for any topic, the word distribution satisfies
βm＝Dirichlet(η → )
Both text-topic, topic-term satisfy a polynomial distribution.
The posterior probability of the topic satisfies
The posterior probability of a word-topic satisfies:
the joint probability is as follows:
in general, the Gibbs sampling algorithm is used to solve the LDA
The key of solving the Gibbs Sampling algorithm is to solve the condition distribution corresponding to the ith word in the corpus
Assuming the word wi=t, then
Since only conjugated structures are involved in the above
Thus there is
The formula of Gibbs Samplin for LDA is then
The flow of the LDA Gibbs sampling algorithm is in fact:
and randomly assigning a topic to each word of the text, recalculating and updating the topic of each word according to the Gibbs sampling algorithm, and repeating the process until the model converges, wherein the topic of the obtained word is the topic distribution result of the text.
Through template matching, the program can identify more spoken search text or ambiguous text, accurately hit the most concerned problems of the user, and avoid the process of manually screening results by the user.
Step 50 makes a more extensive intent recognition based on TextCNN as a result of the open information on the basis of the known word vector;
textCNN is a deep learning algorithm for text classification, which is proposed on the basis of CNN (convolutional neural network), and compared with CNN, textCNN takes a word vector array as input, so that the original structure is simplified, only one convolutional layer and pooling layer are reserved, and finally classification is realized through softmax. The textCNN has the advantages of simple structure and few parameters, so that the textCNN has a faster training speed, and the textCNN can be used for reducing maintenance cost under the condition of more data updating iterations. According to the behavior statistics of the user, such as browsing records, operation logs and the like, search preference of the user is obtained as a classification standard, and the obtained result can be more close to the actual use requirement of the user.
Step 60 extracts keywords from the domain map.
Through the steps, the program extracts the topics possibly contained in the text, but in the actual data, the description of the topics may have various methods, namely one topic may correspond to a plurality of entities or the topics are contained in a certain entity, so that in order to obtain the user intention more accurately, the possible context word association words need to be matched in a previously constructed domain map, the intention of a deeper layer of the user is mined, the sub-domain associated with the keywords in the search statement is explored, for example, the user inputs a word "Beijing expressway", in the map, "great and high speed", "Beijing and high speed", and the like can be returned to the user as association results, or when the user searches for "Beijing and high speed", the information such as "length", "speed limit", "highway grade", "expressway entrance", and the like is the potential intention of the user.
For accurate results, input of, for example, "items with design speed greater than 80" to obtain an accurate answer is shown in fig. 2: it can be seen that the terms such as "six lines", "road base width" have been precisely located by the action of the domain map to the names of the data stores in the database, and the answers of the user search sentences are precisely provided.
For multi-purpose recognition, inputs such as "which of items have a design speed greater than 80 and an investment amount greater than 100 ten thousand" are shown in fig. 3, it can be clearly seen from the figure that information including a retrieval intention is accurately retrieved.
The following is a spoken description of what items are, for example, "how many items are in total in Guizhou". "As can be seen in FIG. 4, the road specific information in all Guizhou provinces below and the results of satisfying the question are shown on the page.
The embodiment integrates the advantages of the traditional syntactic dependency analysis and the part-of-speech analysis, combines the field subject word list, quickly matches the field professional words, analyzes the sentence structure, forms question-answer type retrieval conditions on the basis of obtaining the retrieval text subject, and accurately hits the question answers of the user.
The method combines the traditional method (syntactic dependency and part-of-speech analysis) and the deep learning algorithm (similarity calculation with templates is introduced as possible subjects of sentence division, and search sentences are classified by combining a textCNN algorithm), and the results are combined and preferentially used, so that the accuracy problem hit rate is higher, and the openness problem hit result is closer to the potential intention of a user.
Although the embodiments of the present invention are described above, the embodiments are only used for facilitating understanding of the present invention, and are not intended to limit the present invention. Any person skilled in the art can make any modification and variation in form and detail without departing from the spirit and scope of the present disclosure, but the scope of the present disclosure is still subject to the scope of the appended claims.
Claims (2)
1. A knowledge graph-based question and answer intention recognition method, the method comprising:
step A, constructing a domain topic dictionary;
b, constructing a template;
step C, performing part-of-speech analysis and syntactic dependency analysis on the search statement;
step D, calculating the similarity between the search text and the template sample by using a word vector and an LDA algorithm;
step E, based on the known word vector, performing broader intention recognition based on the textCNN to serve as a result of the open information;
step F, extracting keywords from the domain map;
in the step C: on the basis of obtaining the dependency relationship and the part of speech of the words, the relationship of the words is required to be specifically analyzed, and more topics are matched by combining with the constructed domain topic dictionary;
the part of speech of If word is time, place name, person name:
filling corresponding slots
If has a master-predicate-guest relationship
If subject, object to topic dictionary matching
Judging specific retrieval operation according to predicates, and generating a conditional statement together with the object of the subject
Else
Open matching as keywords into the map
Elif has a centering relationship and an in-form relationship
If fixed language, scholarly language and theme dictionary matching
Performing search operation by using the fixed language and the scholarly language as limiting conditions
Elif's idiom, a possible aggregation operation
Conditional definition of keywords for fixed language and scholarly modified language
Else
Performing open matching as keywords in the map;
in step D: generating word vectors by professional data and general data in the field through word segmentation, converting words into dense vectors and judging the similarity of the words according to the distance between the word vectors by using the text of word stopping work; the two sentences have no obvious relation on the composition words, so LDA is needed to be introduced to help the program to better understand possible relations among the sentences; the LDA is based on a Bayesian model, and the probability is gradually updated according to the obtained posterior distribution in the training process on the premise of adopting the prior probability; presetting a possible topic number in a home page in an LDA model, after the assumed topic number is provided, assuming that the prior distribution of texts accords with the Dirichlet distribution, and setting words in the texts to accord with the Dirichlet distribution, wherein the text-topic and topic-word all meet a polynomial distribution;
solving the LDA by employing a Gibbs sampling algorithm includes:
randomly assigning a topic to each word of the text, recalculating and updating the topic of each word according to the Gibbs sampling algorithm, and repeating the process until the model converges, wherein the topic of the obtained word is the topic distribution result of the text;
through template matching, the program identifies more spoken search text or ambiguous text, accurately hits the most concerned problems of the user, and avoids the process of manually screening results by the user;
the textCNN is a deep learning algorithm for text classification, which is provided on the basis of a convolutional neural network CNN, takes a word vector array as input, simplifies the original structure, only keeps one convolutional layer and pooling layer, and finally realizes classification through softmax;
the theme contained in the text is extracted in the step F, and one theme in the theme corresponds to a plurality of entities or the theme is contained in a certain entity; matching possible related words of the upper and lower position words in a pre-constructed domain map, mining deeper intention of a user, and exploring sub-domains related to keywords in a search sentence.
2. The knowledge-based question and answer intent recognition method of claim 1, wherein the syntactic relationship in the step C includes a main-predicate relationship, a move-guest relationship, a centering relationship, a mid-state relationship and a parallel relationship.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010068158.8A CN111291156B (en) | 2020-01-21 | 2020-01-21 | Knowledge graph-based question and answer intention recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010068158.8A CN111291156B (en) | 2020-01-21 | 2020-01-21 | Knowledge graph-based question and answer intention recognition method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111291156A CN111291156A (en) | 2020-06-16 |
| CN111291156B true CN111291156B (en) | 2024-01-12 |
Family
ID=71021315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010068158.8A Active CN111291156B (en) | 2020-01-21 | 2020-01-21 | Knowledge graph-based question and answer intention recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111291156B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111782769B (en) * | 2020-07-01 | 2022-07-08 | 重庆邮电大学 | Intelligent knowledge graph question-answering method based on relation prediction |
| CN112100405B (en) * | 2020-09-23 | 2024-01-30 | 中国农业大学 | Veterinary drug residue knowledge graph construction method based on weighted LDA |
| CN112270166A (en) * | 2020-10-26 | 2021-01-26 | 摩森特(北京)科技有限公司 | Method for quickly making and creating 5G message |
| CN112905774A (en) * | 2021-02-22 | 2021-06-04 | 武汉市聚联科软件有限公司 | Human-computer conversation deep intention understanding method based on affair map |
| CN113127627B (en) * | 2021-04-23 | 2023-01-17 | 中国石油大学(华东) | Poetry recommendation method based on LDA theme model and poetry knowledge map |
| CN113420130A (en) * | 2021-05-12 | 2021-09-21 | 同方知网数字出版技术股份有限公司 | Intention recognition algorithm based on government affair question-answering system |
| CN113239163A (en) * | 2021-05-12 | 2021-08-10 | 同方知网数字出版技术股份有限公司 | Intelligent question-answering method and system based on traffic big data |
| CN113051374B (en) * | 2021-06-02 | 2021-08-31 | 北京沃丰时代数据科技有限公司 | Text matching optimization method and device |
| CN113590797A (en) * | 2021-08-05 | 2021-11-02 | 云上贵州大数据产业发展有限公司 | Intelligent operation and maintenance customer service system and implementation method |
| CN113792116B (en) * | 2021-08-25 | 2024-03-29 | 北京库睿科技有限公司 | Multi-vertical-domain multi-intention hierarchical judgment method and system based on search word semantics |
| CN117151107A (en) * | 2022-05-24 | 2023-12-01 | 中国电信股份有限公司 | Intention recognition method and device, storage medium and electronic equipment |
| CN115809321B (en) * | 2022-10-09 | 2023-08-04 | 中原工学院 | Retrieval system and method based on knowledge graph |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103823848A (en) * | 2014-02-11 | 2014-05-28 | 浙江大学 | LDA (latent dirichlet allocation) and VSM (vector space model) based similar Chinese herb literature recommendation method |
| CN107562831A (en) * | 2017-08-23 | 2018-01-09 | 中国软件与技术服务股份有限公司 | A kind of accurate lookup method based on full-text search |
| CN109213999A (en) * | 2018-08-20 | 2019-01-15 | 成都佳发安泰教育科技股份有限公司 | A kind of subjective item methods of marking |
| CN109492077A (en) * | 2018-09-29 | 2019-03-19 | 北明智通（北京）科技有限公司 | The petrochemical field answering method and system of knowledge based map |
| CN109635117A (en) * | 2018-12-26 | 2019-04-16 | 零犀(北京)科技有限公司 | A kind of knowledge based spectrum recognition user intention method and device |
| CN109885660A (en) * | 2019-02-22 | 2019-06-14 | 上海乐言信息科技有限公司 | A kind of question answering system and method based on information retrieval that knowledge mapping is energized |
| WO2019144926A1 (en) * | 2018-01-26 | 2019-08-01 | 上海智臻智能网络科技股份有限公司 | Intelligent interaction method and apparatus, computer device and computer-readable storage medium |
| CN110377911A (en) * | 2019-07-23 | 2019-10-25 | 中国工商银行股份有限公司 | Intension recognizing method and device under dialogue frame |
| CN110399457A (en) * | 2019-07-01 | 2019-11-01 | 吉林大学 | A kind of intelligent answer method and system |
| CN110633724A (en) * | 2018-06-25 | 2019-12-31 | 中兴通讯股份有限公司 | Intention recognition model dynamic training method, device, equipment and storage medium |
-
2020
- 2020-01-21 CN CN202010068158.8A patent/CN111291156B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103823848A (en) * | 2014-02-11 | 2014-05-28 | 浙江大学 | LDA (latent dirichlet allocation) and VSM (vector space model) based similar Chinese herb literature recommendation method |
| CN107562831A (en) * | 2017-08-23 | 2018-01-09 | 中国软件与技术服务股份有限公司 | A kind of accurate lookup method based on full-text search |
| WO2019144926A1 (en) * | 2018-01-26 | 2019-08-01 | 上海智臻智能网络科技股份有限公司 | Intelligent interaction method and apparatus, computer device and computer-readable storage medium |
| CN110633724A (en) * | 2018-06-25 | 2019-12-31 | 中兴通讯股份有限公司 | Intention recognition model dynamic training method, device, equipment and storage medium |
| CN109213999A (en) * | 2018-08-20 | 2019-01-15 | 成都佳发安泰教育科技股份有限公司 | A kind of subjective item methods of marking |
| CN109492077A (en) * | 2018-09-29 | 2019-03-19 | 北明智通（北京）科技有限公司 | The petrochemical field answering method and system of knowledge based map |
| CN109635117A (en) * | 2018-12-26 | 2019-04-16 | 零犀(北京)科技有限公司 | A kind of knowledge based spectrum recognition user intention method and device |
| CN109885660A (en) * | 2019-02-22 | 2019-06-14 | 上海乐言信息科技有限公司 | A kind of question answering system and method based on information retrieval that knowledge mapping is energized |
| CN110399457A (en) * | 2019-07-01 | 2019-11-01 | 吉林大学 | A kind of intelligent answer method and system |
| CN110377911A (en) * | 2019-07-23 | 2019-10-25 | 中国工商银行股份有限公司 | Intension recognizing method and device under dialogue frame |
Non-Patent Citations (4)
| Title |
|---|
| "基于主题模型的文本相似度计算研究与实现";孙昌年;《中国优秀硕士学位论文全文数据库》;20121031;全文 * |
| "结合主题模型词向量的CNN文本分类";牛雪莹;《计算机与现代化》;20191029;全文 * |
| A Survey on Application of Knowledge Series;Xiaohan Zou;Journal of Physics;全文 * |
| Introduction of Neural Network based Approaches for Question Answering over Knowledge Graphs;Nilesh Chakraborty et al.;arXiv;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111291156A (en) | 2020-06-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111291156B (en) | Knowledge graph-based question and answer intention recognition method | |
| CN109271529B (en) | Method for constructing bilingual knowledge graph of Xilier Mongolian and traditional Mongolian | |
| CN108304375B (en) | Information identification method and equipment, storage medium and terminal thereof | |
| CN107798140B (en) | Dialog system construction method, semantic controlled response method and device | |
| CN111931506B (en) | Entity relationship extraction method based on graph information enhancement | |
| KR20190133931A (en) | Method to response based on sentence paraphrase recognition for a dialog system | |
| CN112069298A (en) | Human-computer interaction method, device and medium based on semantic web and intention recognition | |
| CN109635288A (en) | A kind of resume abstracting method based on deep neural network | |
| CN108874896B (en) | Humor identification method based on neural network and humor characteristics | |
| US20230069935A1 (en) | Dialog system answering method based on sentence paraphrase recognition | |
| CN113282711B (en) | Internet of vehicles text matching method and device, electronic equipment and storage medium | |
| CN112434164B (en) | Network public opinion analysis method and system taking topic discovery and emotion analysis into consideration | |
| CN111324691A (en) | Intelligent question-answering method for minority nationality field based on knowledge graph | |
| CN110717045A (en) | Letter element automatic extraction method based on letter overview | |
| CN113505209A (en) | Intelligent question-answering system for automobile field | |
| CN115146629A (en) | News text and comment correlation analysis method based on comparative learning | |
| CN112733547A (en) | Chinese question semantic understanding method by utilizing semantic dependency analysis | |
| CN114417851A (en) | Emotion analysis method based on keyword weighted information | |
| CN114491062B (en) | Short text classification method integrating knowledge graph and topic model | |
| CN115712700A (en) | Hot word extraction method, system, computer device and storage medium | |
| CN114064901B (en) | Book comment text classification method based on knowledge graph word meaning disambiguation | |
| CN111178080A (en) | Named entity identification method and system based on structured information | |
| CN114841353A (en) | Quantum language model modeling system fusing syntactic information and application thereof | |
| CN112380848B (en) | Text generation method, device, equipment and storage medium | |
| CN113157887A (en) | Knowledge question-answering intention identification method and device and computer equipment |
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 |