CN107480125B - Relation linking method based on knowledge graph - Google Patents

Relation linking method based on knowledge graph Download PDF

Info

Publication number
CN107480125B
CN107480125B CN201710543849.7A CN201710543849A CN107480125B CN 107480125 B CN107480125 B CN 107480125B CN 201710543849 A CN201710543849 A CN 201710543849A CN 107480125 B CN107480125 B CN 107480125B
Authority
CN
China
Prior art keywords
relation
text
knowledge
cluster
graph
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
Application number
CN201710543849.7A
Other languages
Chinese (zh)
Other versions
CN107480125A (en
Inventor
李智星
杨茜
任诗雅
沈柯
李苑
王国胤
胡峰
王进
雷大江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN201710543849.7A priority Critical patent/CN107480125B/en
Publication of CN107480125A publication Critical patent/CN107480125A/en
Application granted granted Critical
Publication of CN107480125B publication Critical patent/CN107480125B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/205Parsing
    • G06F40/211Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35Clustering; Classification
    • G06F16/355Class or cluster creation or modification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/30Semantic analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Machine Translation (AREA)

Abstract

The invention provides a relation linking method based on a knowledge graph, which comprises the steps of firstly, utilizing SparQ L query sentences to find a triple list set containing a certain relation, matching relation texts from unstructured texts, utilizing a L SWMD algorithm to obtain a similarity matrix of the relation texts, utilizing a density peak value clustering algorithm to cluster the relation texts to obtain a relation text cluster, extracting the positions of all words in the cluster based on the relation text cluster, utilizing beta distribution to perform fitting to obtain a word distribution mode of the relation text cluster, utilizing a word distribution mode to construct vectors for candidate relation texts with undetermined relation in an unstructured text in an open field, utilizing a GBDT classifier to perform recognition, and further utilizing the GBDT classifier to link with the relation in the knowledge graph.

Description

Relation linking method based on knowledge graph
Technical Field
The invention belongs to the field of natural language processing, and particularly relates to a relation linking method based on a knowledge graph.
Background
Exploring and understanding knowledge on the internet is one of the long-term goals in the field of artificial intelligence. With the advent of some distributed systems, storing and utilizing data on the internet has not been a difficult problem, but enabling computers to understand and manipulate human natural language has remained a significant challenge. The birth of the knowledge-graph helps the computer understand the natural language. A knowledge graph is intended to describe various entities or concepts that exist in the real world. Where each entity or concept is identified by a globally uniquely determined ID, referred to as their identifier. Each attribute-value pair is used to characterize an entity's intrinsic properties, while a relationship is used to link two entities, characterizing an association between them. A knowledge graph can also be viewed as a large graph, where nodes represent entities or concepts and edges are composed of attributes or relationships. The popular knowledge maps at present include DBpedia, Wikitata, OpenCyc, YAGO and the like.
Knowledge graphs play a crucial role in many areas, such as semantic search, question-and-answer systems, etc. One of the important difficulties is mapping natural text into a knowledge-graph. Entity linking addresses this difficulty by mapping some entities of natural text into a knowledge-graph. At present, entity linkage is a relatively mature research, but relationship linkage is rarely concerned. Relationship links and relationship extractions are distinct, with the emphasis on relationship extraction being to identify a relationship between two entities, and relationship links attempting to find a textual representation of a target relationship.
A relation link system based on knowledge graph is to use the structured data of knowledge graph to learn the common expression of a certain relation in knowledge graph and to build the common expression model of the relation. When processing the unstructured text, the characters are matched by using the common expression model, so that the unstructured text is mapped to a certain relation in the knowledge graph, the relation link is realized, meanwhile, the deep relation between entities can be inferred, and the knowledge graph is enriched.
Disclosure of Invention
The present invention is directed to solving the above problems of the prior art. The relation linking method based on the knowledge graph can effectively solve the problem that the natural language and the knowledge graph are not linked sufficiently and help a computer to understand the natural language better. The technical scheme of the invention is as follows:
a relation linking method based on knowledge graph includes the following steps:
s1, acquiring the knowledge map and the unstructured text data set, preprocessing the data, and utilizing the knowledge
Labeling the atlas, obtaining a relation text in the unstructured text as a training set:
s2, obtaining relation text-based pairwise word movement distance algorithm based on L SWMD position sensitivity
And (3) clustering the similarity matrix based on the similarity matrix to obtain a relation text cluster:
s3, fitting the positions of the words in the relation text cluster by utilizing the beta distribution to obtain a word distribution mode:
s4, converting the training set into vectors by using a word distribution mode, and training by using a GBDT gradient lifting tree to obtain a classifier:
and S5, matching the unstructured text which is not labeled by the knowledge graph or cannot be labeled by the knowledge graph by using the relational text cluster, judging by using the GBDT classifier, and linking to the corresponding relation of the knowledge graph if the unstructured text is judged to be true.
Further, the step of performing regular noise reduction on the relationship text includes: and screening the relation texts with the lengths of more than 3 and less than 11.
The step S1 of acquiring the knowledge graph data set and preprocessing data to acquire the relational text specifically comprises the steps of obtaining entity pairs from the knowledge graph by using SparQ L and establishing an entity pair list, obtaining corresponding Wikipedia articles according to the subject, using an nltk tool to perform clause segmentation, marking subj if the sentence comprises the subject, the alias of the subject and the main part of the subject, marking obj if the sentence comprises the object, the alias of the object and the main part of the object, and intercepting the character part between the subj and the obj as the relational text.
Further, the step of clustering the relational texts by using the L SWMD algorithm of the step S2 to obtain the relational text cluster includes converting words into vectors by using word2vec, calculating a semantic distance matrix between the relational texts by using the word vectors, calculating a syntactic distance matrix between the relational texts by using positions of the words, taking the sum of the parameter α multiplied by the semantic distance matrix and the (1- α) multiplied by the syntactic distance matrix as the input of the EMD to obtain the similarity between every two relational texts, constructing the similarity matrix as the input, clustering the similarity matrix by using a density peak algorithm to obtain the relational text cluster.
Further, the step S3 of obtaining a word distribution pattern includes the steps of counting positions of words in the relational text, fitting the position information of the words by utilizing beta distribution to obtain a parameter α, then counting the probability γ of the occurrence of the words in the cluster, and representing the words as triples of (α, γ), wherein the triples of all the words in the cluster form the word distribution pattern of the cluster.
Further, the training step of the classifier comprises the steps of initializing vectors according to the size of the class clusters, finding a word sequence with the maximum frequency by sentences in a training set through a sliding window from 4 to 10, calculating the matching degree of words and filling the matching degree of the words and the words in the corresponding positions of the vectors by using α parameters and the positions of the words and the words in the word sequence, splicing the vectors of the word sequence in each class cluster, and training the vectors by using GBDT to obtain the classifier.
Further, the step S5 of determining the unstructured text relationship by using a classifier and linking to the attribute page corresponding to the knowledge graph specifically includes the steps of: and converting the unstructured text into a vector by using a word distribution mode, identifying whether the text contains the relation by using a classifier, and linking the relation text of the text to an attribute page corresponding to the knowledge graph if the text contains the relation.
The invention has the following advantages and beneficial effects:
1. the position-sensitive word movement distance is provided for calculating the similarity of the relational text, the semantic distance and the grammatical distance can be considered at the same time, and a better similarity calculation result is obtained.
2. A word distribution bag model is provided, and the relation text cluster is represented as beta distribution of a series of words, so that the data volume of the text relation cluster is greatly reduced, and the frequency and position information of the words in the cluster are kept.
3. A new vector representation mode of the relation text is provided, for each relation text class cluster, one relation text is represented as a vector formed by words contained in the relation text class cluster, and for a plurality of relation text class clusters, one relation text is represented as the splicing of a plurality of corresponding vectors. The method can effectively express the association relationship between the relationship text and the class cluster.
4. A relation link framework based on the knowledge graph is provided, word sequences in unstructured natural texts can be linked to relations in the knowledge graph, and a computer can be helped to understand natural languages better.
Drawings
FIG. 1 is an overall flow chart of the preferred embodiment of the present invention;
FIG. 2 is a flow chart of data collection and preprocessing.
Detailed Description
The technical solutions in the embodiments of the present invention will be described in detail and clearly with reference to the accompanying drawings. The described embodiments are only some of the embodiments of the present invention.
The technical scheme for solving the technical problems is as follows:
the main design concept of the technical scheme of the invention is to design an automatic linking algorithm for the relation in the knowledge graph. And training a common expression model of a certain relation by using the structured data of the knowledge graph. And matching and identifying the common expressions and the unstructured texts, and mapping the successfully matched unstructured texts to the corresponding relation in the knowledge graph, so that relation linkage is realized, and the problem of relation mapping between the unstructured natural language texts and the structured knowledge graph is solved. Meanwhile, the deep relation between the entities can be inferred, and the knowledge map is enriched.
The following describes the technical solution of the present invention in further detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an embodiment of a relation linking system based on knowledge-graph according to the present invention is shown, which is mainly implemented as follows:
and step S1, collecting and preprocessing data to obtain corresponding relation texts.
Step S2, clustering the relation texts by using L SWMD algorithm to obtain relation text cluster, wherein the specific process is as follows:
the invention improves the characteristic that the WMD can not process the syntactic structure information, namely L fact Sensitive Word generators' Distance (L SWMD). A relational text is used as { w1,w2,…wnDenotes, wiThe position of (c) can be calculated by the following formula:
Figure BDA0001342552370000051
after the position is added, the formula of the WMD is as follows:
Figure BDA0001342552370000052
Figure BDA0001342552370000053
Figure BDA0001342552370000054
wherein s isiFor the ith word in the relational text, diIs the weight of the ith word in the sentence, T is siTo sjD is siTo sjSemantic and syntax based distance matrices.
The detailed formula of the distance matrix is as follows:
Dsem+loc(si,s′j)=αDsem(si,s′j)+(1-α)Dloc(si,s′j),α∈[0,1]
where α is a hyperparameter.
The similarity matrix obtained by the method is clustered by using a density peak algorithm, and the similarity matrix can be used for clustering the texts expressing similar relations together to obtain a common relation text cluster.
Step S3, fitting the positions of the words by utilizing the beta distribution to obtain a word distribution mode, wherein the specific process is as follows:
the method comprises the steps of obtaining a relation text cluster, calculating positions of words in the relation text cluster by using the relation text cluster obtained in the last step, fitting position information of the words by using beta distribution to obtain a parameter α, then calculating the probability gamma of the words appearing in the relation text cluster, and representing the words as (α, gamma) triples, wherein the triples of all the words in the relation text cluster form a word distribution mode of the relation text cluster.
Step S4, converting the training set into vectors by using a word distribution mode, and training the classifier by using GBDT, wherein the specific process is as follows:
in the training process, a sliding window of 4 to 10 is used, and the frequency in the word distribution pattern is used, so that the candidate relation sequence with the largest sum of the word sequence frequencies in the window is selected. And then, calculating the matching degree of each word by using the word distribution mode and the positions of the words in the candidate relation sequence so as to obtain a matching degree vector of the sentence, splicing the vectors of the word sequence in each class cluster, and training the vectors by using a classifier so as to obtain a classifier.
In step S5, the newly input unstructured text may be converted into a vector by using a word distribution pattern, and the classifier trained in step S4 may be used to identify whether the text contains the relationship. And if so, linking the relation text of the text to an attribute page corresponding to the knowledge graph.
In the present invention, please refer to fig. 2 for data collection and pre-processing, which is a flow chart of data collection and pre-processing according to the present invention, and the main implementation process is as follows:
step S6, obtain entity pairs from the knowledge graph using SparQ L, and build entity pair list.
And step S7, obtaining a corresponding Wikipedia article according to the subject.
In step S8, clauses are divided by the nltk tool.
In step S9, if the sentence includes the subject, the alias of the subject, and the main body part of the subject, it is marked as subj.
In step S10, if the sentence includes the object, the alias of the object, and the body part of the object, it is marked as obj.
And step S11, further processing the sentence, and intercepting the character part between subj and obj as a relation text.
And step S12, screening the relation texts with the length being more than 3 and less than 11.
In summary, the method includes the steps of finding a relation text containing a certain relation by using data of a knowledge graph, obtaining a similarity matrix of the relation text by using L SWMD algorithm, obtaining a relation text cluster by using density peak value clustering algorithm, extracting positions of all words in the cluster based on the relation text cluster, fitting by using beta distribution to obtain a word distribution mode of the relation text cluster, and for an unstructured natural text, constructing a vector through the word distribution mode, identifying whether the text contains the relation through GBDT, and further linking the text with the relation in the knowledge graph.
The above examples are to be construed as merely illustrative and not limitative of the remainder of the disclosure. After reading the description of the invention, the skilled person can make various changes or modifications to the invention, and these equivalent changes and modifications also fall into the scope of the invention defined by the claims.

Claims (6)

1. A relation linking method based on knowledge graph is characterized by comprising the following steps:
s1, acquiring a knowledge graph and an unstructured text data set, preprocessing the data, labeling by using the knowledge graph, and acquiring a relation text in the unstructured text as a training set;
s2, obtaining a similarity matrix based on pairwise relation texts by adopting a L SWMD position-sensitive word moving distance algorithm, and clustering based on the similarity matrix to obtain a relation text cluster, wherein the step of clustering the relation texts by using the L SWMD algorithm of the step S2 to obtain the relation text cluster comprises the steps of converting words into vectors by using word2vec, calculating a semantic distance matrix between the relation texts by using the word vectors, calculating a syntactic distance matrix between the relation texts by using the positions of the words, multiplying a parameter α by the sum of the semantic distance matrix and (1- α) by the syntactic distance matrix as the input of EMD, taking α as a hyper parameter, obtaining the similarity based on the relation texts in a value range of [0, 1], constructing the similarity matrix as the input, and obtaining the relation text cluster by using a density peak value clustering algorithm;
s3, fitting the positions of the words in the relation text cluster by utilizing the Beta distribution to obtain a word distribution pattern, wherein the step S3 of obtaining the word distribution pattern comprises the steps of counting the positions of the words in the relation text, fitting the position information of the words by utilizing the Beta distribution to obtain a parameter α, then counting the probability gamma of the words appearing in the cluster, and representing the words as (α, gamma) triples;
s4, converting the training set into vectors by using a word distribution mode, wherein the relation text labeled by using the knowledge graph is the training set, and training by using a GBDT gradient lifting tree to obtain a classifier;
and S5, matching the unstructured text which is not labeled by the knowledge graph or cannot be labeled by the knowledge graph by using the relational text cluster, judging by using the classifier, and linking to the corresponding relation of the knowledge graph if the unstructured text is judged to be true.
2. The relation linking method based on the knowledge-graph as claimed in claim 1, further comprising the step of performing regular denoising on the relation text after the relation text is obtained.
3. The relation linking method based on the knowledge-graph as claimed in claim 2, wherein the step of regular denoising the relation text comprises: and screening the relation texts with the lengths of more than 3 and less than 11.
4. The relation linking method based on the knowledge-graph of any one of claims 1 to 3, wherein the step S1 is used for collecting a knowledge-graph data set and preprocessing data to obtain a relation text, and specifically comprises the steps of obtaining an entity pair from the knowledge-graph by using spark Q L, establishing an entity pair list, obtaining a corresponding Wikipedia article according to a subject, carrying out sentence segmentation by using an nltk tool, if the sentence contains the subject, an alias of the subject and a main part of the subject, marking the sentence as subj, if the sentence contains the subject, the alias of the subject and the main part of the object, marking the sentence as obj, and intercepting a character part between the subj and the obj from the sentence as the relation text.
5. The relation linking method based on the knowledge graph of claim 1, wherein the training step of the classifier comprises the steps of initializing a vector according to the size of a class cluster, finding a word sequence with the maximum frequency by using a sliding window of 4 to 10 for sentences in a training set, calculating the matching degree of words and filling the matching degree into the corresponding positions of the vector by using α parameters and the positions of the words in the word sequence, splicing the vectors of the word sequence in each class cluster, and training the vectors by using GBDT to obtain the classifier.
6. The relation linking method based on the knowledge-graph as claimed in claim 1, wherein the step S5 of distinguishing the unstructured text relation by using a classifier, and linking to the attribute page corresponding to the knowledge-graph specifically comprises the steps of: and converting the unstructured text into a vector by using a word distribution mode, identifying whether the text contains the relation by using a classifier, and linking the relation text of the text to an attribute page corresponding to the knowledge graph if the text contains the relation.
CN201710543849.7A 2017-07-05 2017-07-05 Relation linking method based on knowledge graph Active CN107480125B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710543849.7A CN107480125B (en) 2017-07-05 2017-07-05 Relation linking method based on knowledge graph

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710543849.7A CN107480125B (en) 2017-07-05 2017-07-05 Relation linking method based on knowledge graph

Publications (2)

Publication Number Publication Date
CN107480125A CN107480125A (en) 2017-12-15
CN107480125B true CN107480125B (en) 2020-08-04

Family

ID=60595600

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710543849.7A Active CN107480125B (en) 2017-07-05 2017-07-05 Relation linking method based on knowledge graph

Country Status (1)

Country Link
CN (1) CN107480125B (en)

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109993381B (en) * 2017-12-29 2021-11-30 中国移动通信集团湖北有限公司 Demand management application method, device, equipment and medium based on knowledge graph
CN109800317B (en) * 2018-03-19 2022-03-11 中山大学 Image query answering method based on image scene map alignment
US10915577B2 (en) * 2018-03-22 2021-02-09 Adobe Inc. Constructing enterprise-specific knowledge graphs
CN110019843B (en) * 2018-09-30 2020-11-06 北京国双科技有限公司 Knowledge graph processing method and device
CN110147421B (en) * 2019-05-10 2022-06-21 腾讯科技(深圳)有限公司 Target entity linking method, device, equipment and storage medium
CN110489560A (en) * 2019-06-19 2019-11-22 民生科技有限责任公司 The little Wei enterprise portrait generation method and device of knowledge based graphical spectrum technology
CN110909174B (en) * 2019-11-19 2022-01-04 南京航空航天大学 Knowledge graph-based method for improving entity link in simple question answering
CN111259653B (en) * 2020-01-15 2022-06-24 重庆邮电大学 Knowledge graph question-answering method, system and terminal based on entity relationship disambiguation
CN111309922B (en) * 2020-01-19 2023-11-17 清华大学 Map construction method, accident classification device, computer equipment and medium
CN111737456A (en) * 2020-05-15 2020-10-02 恩亿科(北京)数据科技有限公司 Corpus information processing method and apparatus
CN111611405B (en) * 2020-05-22 2023-03-21 北京明略软件系统有限公司 Knowledge graph construction method and device, electronic equipment and storage medium
CN111813914B (en) * 2020-07-13 2021-07-06 龙马智芯(珠海横琴)科技有限公司 Question-answering method and device based on dictionary tree, recognition equipment and readable storage medium
CN111914568B (en) * 2020-07-31 2024-02-06 平安科技(深圳)有限公司 Method, device and equipment for generating text sentence and readable storage medium
CN111709528B (en) * 2020-08-18 2021-01-05 北京工业大数据创新中心有限公司 Expert rule protection method and device
CN112417456B (en) * 2020-11-16 2022-02-08 中国电子科技集团公司第三十研究所 Structured sensitive data reduction detection method based on big data
CN113449084A (en) * 2021-09-01 2021-09-28 中国科学院自动化研究所 Relationship extraction method based on graph convolution
CN113821647B (en) * 2021-11-22 2022-02-22 山东捷瑞数字科技股份有限公司 Construction method and system of knowledge graph in engineering machinery industry
CN117150050B (en) * 2023-10-31 2024-01-26 卓世科技(海南)有限公司 Knowledge graph construction method and system based on large language model

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104035917A (en) * 2014-06-10 2014-09-10 复旦大学 Knowledge graph management method and system based on semantic space mapping
CN106777232A (en) * 2016-12-26 2017-05-31 上海智臻智能网络科技股份有限公司 Question and answer abstracting method, device and terminal
CN106777275A (en) * 2016-12-29 2017-05-31 北京理工大学 Entity attribute and property value extracting method based on many granularity semantic chunks

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104035917A (en) * 2014-06-10 2014-09-10 复旦大学 Knowledge graph management method and system based on semantic space mapping
CN106777232A (en) * 2016-12-26 2017-05-31 上海智臻智能网络科技股份有限公司 Question and answer abstracting method, device and terminal
CN106777275A (en) * 2016-12-29 2017-05-31 北京理工大学 Entity attribute and property value extracting method based on many granularity semantic chunks

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Entity Linking with a Knowledge Base:Issues,Techniques,and Solutions;Wei Shen et al.;《IEEE Transaction on Knowledge and Data Engineering》;20151231;第27卷(第2期);第1-20页 *
From Word Embeddings to Document Distances;Matt J.Kusner et al.;《Proceedings of the 32nd International Conference on Machine Learning》;20151231;第1-10页 *
一种基于网格的密度峰值聚类算法;王飞 等;《小型微型计算机系统》;20170531;第1034-1038页 *

Also Published As

Publication number Publication date
CN107480125A (en) 2017-12-15

Similar Documents

Publication Publication Date Title
CN107480125B (en) Relation linking method based on knowledge graph
CN106874378B (en) Method for constructing knowledge graph based on entity extraction and relation mining of rule model
CN109543183B (en) Multi-label entity-relation combined extraction method based on deep neural network and labeling strategy
CN107944559B (en) Method and system for automatically identifying entity relationship
CN109388795B (en) Named entity recognition method, language recognition method and system
CN106845411B (en) Video description generation method based on deep learning and probability map model
CN109753660B (en) LSTM-based winning bid web page named entity extraction method
CN106682411A (en) Method for converting physical examination diagnostic data into disease label
CN111274790B (en) Chapter-level event embedding method and device based on syntactic dependency graph
CN110413787B (en) Text clustering method, device, terminal and storage medium
CN108664599B (en) Intelligent question-answering method and device, intelligent question-answering server and storage medium
CN111858940B (en) Multi-head attention-based legal case similarity calculation method and system
CN111783394A (en) Training method of event extraction model, event extraction method, system and equipment
CN106909537B (en) One-word polysemous analysis method based on topic model and vector space
CN111400455A (en) Relation detection method of question-answering system based on knowledge graph
CN110188359B (en) Text entity extraction method
CN110362678A (en) A kind of method and apparatus automatically extracting Chinese text keyword
CN112559684A (en) Keyword extraction and information retrieval method
CN111143574A (en) Query and visualization system construction method based on minority culture knowledge graph
CN112163097A (en) Military knowledge graph construction method and system
CN109522396B (en) Knowledge processing method and system for national defense science and technology field
Ren et al. Detecting the scope of negation and speculation in biomedical texts by using recursive neural network
CN108763192B (en) Entity relation extraction method and device for text processing
CN111191031A (en) Entity relation classification method of unstructured text based on WordNet and IDF
Yin et al. Sentiment lexical-augmented convolutional neural networks for sentiment analysis

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