CN111125367A - Multi-character relation extraction method based on multi-level attention mechanism - Google Patents

Multi-character relation extraction method based on multi-level attention mechanism Download PDF

Info

Publication number
CN111125367A
CN111125367A CN201911362557.9A CN201911362557A CN111125367A CN 111125367 A CN111125367 A CN 111125367A CN 201911362557 A CN201911362557 A CN 201911362557A CN 111125367 A CN111125367 A CN 111125367A
Authority
CN
China
Prior art keywords
text
layer
word
vector
entity
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.)
Granted
Application number
CN201911362557.9A
Other languages
Chinese (zh)
Other versions
CN111125367B (en
Inventor
蔡毅
刘宸铄
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN201911362557.9A priority Critical patent/CN111125367B/en
Publication of CN111125367A publication Critical patent/CN111125367A/en
Application granted granted Critical
Publication of CN111125367B publication Critical patent/CN111125367B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06F16/35—Clustering; Classification
    • G06F16/355—Class or cluster creation or modification
    • 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
    • G06F16/353—Clustering; Classification into predefined classes
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Machine Translation (AREA)

Abstract

The invention discloses a multi-character relation extraction method based on a multi-level attention mechanism, which comprises the following steps of: preprocessing the collected text; adopting a remote supervision technology to align and label the named entity of the original character to obtain a text containing the entity and entity description information; performing Chinese word vector training on the obtained text containing the entity; constructing an attention mechanism two-way long-time and short-time memory network comprising two levels, and training the constructed model to obtain a multi-classification model for extracting various character relations; and inputting the preprocessed text to obtain a text relation extraction result. The invention overcomes the defects of relation extraction of the existing multiple character relation texts and improves the relation extraction experimental results of the multiple character relation texts.

Description

Multi-character relation extraction method based on multi-level attention mechanism
Technical Field
The invention relates to the field of natural language processing, in particular to a method for extracting relationships among multiple characters based on a multi-level attention mechanism.
Background
With the rapid development of internet technology, text information data in a network grows exponentially, but text information data is often unstructured information. Information extraction is a task of natural language processing, which aims to extract structured information from unstructured text. Information extraction includes two aspects: mingmei Shishi (named as Shisui)A body identification task and a relation extraction task, wherein the former is used for discovering the entities existing in the text, and the latter judges the relation between the discovered entities, namely obtaining the entity pair e for the specified text1And e2And a relation r between the two1，r，e2). The relation extraction task has been widely used in fields of knowledge graph, information retrieval, and the like.
Conventional non-deep learning methods for relationship extraction are typically supervised learning and can be classified into feature-based methods and kernel-based methods, both of which use existing NLP tools and can result in downstream error accumulation. In the deep learning era, a manual feature acquisition mode is avoided, but supervised deep learning requires a large amount of training data to learn features. The labeling of the training data takes a lot of time and effort, and is biased to a certain fixed domain. Mintz et al, 2009, proposed a remote surveillance method that strongly assumes that the entity relationships in the knowledge base are the entity relationships in the text, and generates a large amount of data by aligning the knowledge base with the text.
However, the strong assumption of remote supervision is not necessarily true, and the entity relationships existing in the text are not necessarily the same as those in the knowledge base. To alleviate this drawback, Riedel uses multi-instance learning. Lin first used a segmented convolutional neural network and a sentence-level attention mechanism in 2016, and the use of deep learning and the attention mechanism was introduced to enable the relationship extraction to achieve a better effect.
So far, most of the relation extraction tasks are for English texts, and for Chinese texts, especially for Chinese texts containing various character relations, a method for realizing a better Chinese text multiple character relation extraction method by using a deep learning fusion attention mechanism is urgently needed to be researched.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method for extracting relationships among multiple characters based on a multi-level attention mechanism. The method comprises the steps of obtaining global feature representation of a text by adopting a bidirectional long-time memory network and a word-level attention mechanism, strengthening the weight of a vocabulary which is more important for relation extraction, then obtaining package representation consisting of a plurality of sentence representations by the sentence-level attention mechanism in a multi-instance learning mode, and adding description information of a named entity to strengthen the result of the package representation. The invention obtains better experimental results on the remote supervision relation extraction data set.
The invention can be realized by the following technical scheme:
a multi-character relationship extraction method based on a multi-level attention mechanism comprises the following steps:
preprocessing the collected text;
adopting a remote supervision technology to align and label the named entity of the original character to obtain a text containing the entity and entity description information;
performing Chinese word vector training on the obtained text containing the entity;
constructing an attention mechanism two-way long-time and short-time memory network comprising two levels, and training the constructed model to obtain a multi-classification model for extracting various character relations;
and inputting the preprocessed text to obtain a text relation extraction result.
Specifically, the pretreatment comprises:
removing English data in the text;
removing emoticons and hyperlinks in the text;
removing stop words in the text according to the Chinese stop word list;
and performing Chinese word segmentation on the processed text.
Specifically, in the step of aligning and labeling the original character named entity by adopting the remote monitoring technology, a character name entry is obtained by utilizing Chinese online encyclopedia, two related characters and the relation of the characters form a triple, and finally a character relation knowledge base is constructed. The pair of entities that exist with the knowledge base, i.e., the relationship of two entities, appear in the text as a relationship in a triple. The final annotated data set in the present invention has 35 relationship types.
Specifically, in the step of training the Chinese Word vector of the text, a distributed Word vector representation method Word2Vec is adopted, and the dimension of the output Word vector is set to 300.
Specifically, in the step of constructing the two-level attention mechanism bidirectional long-short time memory network, a birstm (bidirectional long-short time memory network) and two-level attention mechanism network structures are constructed by using a pytorch, wherein the first layer of the network is an embedded layer, the second layer is a bidirectional LSTM layer, the third layer is a word-level attention layer, the fourth layer is a sentence-level attention layer, and the fifth layer is a classifier soffmax layer.
Furthermore, the input of the embedding layer is a trained Word vector sequence, the length of a text sequence (the number of the Word vector sequences) is set to be m, the length of the text sequence is filled with 0 when the length of the text sequence is less than m, the length of the text sequence exceeds m, the length of the relative position of a Word relative to two entities is also m, the embedding layer adopts Baidu encyclopedia as corpus, and Word2Vec is obtained by using a Gensim tool. The dimension of the word vector is dw, the dimension of the randomly initialized position vector is dp, and a vector sequence w is obtained by the method1，w2，…，wm}，wi∈RdWherein d ═ dw + dp × 2.
Further, in the bidirectional LSTM layer, its unidirectional LSTM is expressed as follows:
itto the input gate, ftTo forget the door, ctIn a cellular state, otTo the output gate, htFor the hidden vector, Wx，Wh，WcIn order to be the weight, the weight is,
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+ittanh(Wxcxt+Whcht-1+bc)
ot＝σ(Wxoxt+Whoht-1+Wcoct+bo)
ht＝ottanh(ct)
vector h of bidirectional LSTMtOutput from the forward network
Figure BDA0002337580090000041
And reverse output
Figure BDA0002337580090000042
Are obtained together.
Further, the word-level attention tier is used to enforce weights for the more important words for relationship extraction.
In the attention level of the word level, uiRepresenting the score of the correlation of each word in a sentence, r representing a random query vector, hiIs htEach term of (2), namely each word hidden vector, has the following specific relationship:
ui＝hi·r
αiis the weight obtained by the word-level attention mechanism, and the calculation formula is as follows:
Figure BDA0002337580090000043
s is a vector representation of a sentence, and the calculation formula is as follows:
Figure BDA0002337580090000044
furthermore, the sentence-level attention layer is used for adding entity description information.
In the sentence-level attention layer, eiRepresenting an input sentence siRelation to prediction rkThe calculation formula of the matching degree is as follows:
Figure BDA0002337580090000045
αiis the weight obtained by the sentence level attention mechanism, and the calculation formula is as follows:
Figure BDA0002337580090000051
b is a vector representation of a packet, which is equal to the sum of the weights of all sentences, and the calculation formula is as follows:
Figure BDA0002337580090000052
the obtained packet represents description information of the entity on the splice, i.e. a category information vector of the entity, and is represented as:
Figure BDA0002337580090000053
and the fifth layer is a classifier softmax layer for generating relational extraction results through a softmax multi-classifier.
The method obtains the global feature representation of a text by using a BilSTM (bidirectional long and short term memory network) and a Word level attention (Word level attention mechanism), wherein the Word level attention mechanism is used for strengthening the weight of a vocabulary which is more important for relation extraction, and then a multi-instance learning mode is adopted, and a sequence level attention (Sentence level attention mechanism) obtains a packet representation consisting of a plurality of Sentence representations and adds description information of a named entity to strengthen the result of the packet representation.
Compared with the prior art, the invention has the following beneficial effects:
the invention adopts two-level attention mechanism for Chinese texts with various task relationships, better avoids noise caused by remote supervision, and simultaneously adds entity description information into the Chinese texts, so that the semantic characteristics of the texts are enhanced, and a better relationship extraction result is obtained.
Drawings
FIG. 1 is a flow chart of a method for extracting relationships between multiple people based on a multi-level attention mechanism according to the present invention.
FIG. 2 is a diagram of a multi-level attention mechanism-based multi-person relationship extraction network model according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Examples
Fig. 1 is a flowchart illustrating a method for extracting relationships among multiple persons based on a multi-level attention mechanism, the method including the steps of:
(1) preprocessing the collected text;
in this embodiment, a network-exposed remote-supervised multi-person relationship extraction dataset (e.g., CCKS 2019IPER dataset) is used. The following operations are carried out: firstly, removing English data in a text;
removing special symbols in the text such as: emoji and hyperlinks, representing emoji as "expression", removing hyperlinks, etc.; and removing stop words in the text according to the Chinese stop word list.
(2) Adopting a remote supervision technology to align and label the named entity of the original character to obtain a text containing the entity and entity description information;
in the embodiment, in the entity labeling stage, the name entry of a Chinese character is acquired in an online encyclopedia, two related characters and the relationship of the two related characters form a triple, and finally a character relationship knowledge base is constructed. The pair of entities that exist with the knowledge base, i.e., the relationship of two entities, appear in the text as a relationship in a triple. There are 35 relationship types for the final labeled data set.
(3) Performing Chinese word vector training on the obtained text containing the entity;
in the text vectorization step of this embodiment, the word2vec method is used, which performs chinese word segmentation on the text subjected to the processing by using a word segmentation tool, and performs word2vec training by using a genim packet, so that the vector dimension of each word is 300.
(4) Constructing an attention mechanism two-way long-time and short-time memory network comprising two levels, and training the constructed model to obtain a multi-classification model for extracting various character relations;
as shown in fig. 2, the network model constructed in this embodiment has a structure that includes: an embedding layer, a bidirectional LSTM layer, a word level attention layer, a sentence level attention layer and a softmax classification layer.
The neural network model constructed by the embodiment is trained through a downloaded Chinese Baidu encyclopedia data set, Crossentpy is used as a loss function, Adam is used as an optimization method, training is completed when 15 epochs are trained or loss is not changed in 1000 lots through adjustment of other parameters of the model, a test set is tested, a result obtained through measuring relation extraction by using a P-R curve is obtained, the P-R curve utilizes precision and recall of the result to depict curves, and the upper curve in a two-dimensional coordinate system represents a better relation extraction effect.
(5) And preprocessing the text needing relationship extraction, and inputting the preprocessed text into the trained model to obtain a text relationship extraction result.
The method for extracting the relationships of various characters based on the multi-level attention mechanism, which is established in the method, converts an input text into a vector form through an embedded layer, obtains a hidden vector with more characteristics through a BilSTM layer, extracts more important words from the relationship in the text and obtains higher weight, and the sentence-level attention mechanism is better represented by a packet, so that noise caused by remote supervision can be eliminated, and a better experimental result is obtained.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims (9)

1. A multi-character relation extraction method based on a multi-level attention mechanism is characterized by comprising the following steps:
preprocessing the collected text;
adopting a remote supervision technology to align and label the named entity of the original character to obtain a text containing the entity and entity description information;
performing Chinese word vector training on the obtained text containing the entity;
constructing an attention mechanism two-way long-time and short-time memory network comprising two levels, and training the constructed model to obtain a multi-classification model for extracting various character relations;
and inputting the preprocessed text to obtain a text relation extraction result.
2. The method of claim 1, wherein the pre-processing comprises:
removing English data in the text;
removing emoticons and hyperlinks in the text;
removing stop words in the text according to the Chinese stop word list;
and performing Chinese word segmentation on the processed text.
3. The method as claimed in claim 1, wherein in the step of performing aligned labeling on the named entities of the original characters by using a remote supervision technology, the obtained name entry of the online encyclopedia in chinese is used to combine two characters with relationships and the relationships of the characters into a triple, and finally a character relationship knowledge base is constructed. The pair of entities that exist with the knowledge base, i.e., the relationship of two entities, appear in the text as a relationship in a triple.
4. The method of claim 1, wherein in the step of training the Chinese Word vector on the text, a distributed Word vector representation method Word2Vec is adopted, and the dimension of the output Word vector is set to 300.
5. The method of claim 1, wherein in the step of constructing the bidirectional long-term memory network with two levels of attention mechanisms, the birstm and the two levels of attention mechanism network structures are constructed by using a pytorch, wherein the first layer of the network is an embedding layer, the second layer is a bidirectional LSTM layer, the third layer is a word level attention layer, the fourth layer is a sentence level attention layer, and the fifth layer is a classifier softmax layer.
6. The method of claim 5, wherein the input of the embedding layer is a trained word vector sequence, the length of the text sequence (number of word vector sequences) is set to m, less than m is filled with 0, more than m is truncated, the length of the relative position of a word with respect to two entities in each text is also m, the embedding layer adopts a pre-trained word vector dimension dw and a randomly initialized position vector dimension dp, thereby obtaining a vector sequence w ═ { w ═ w [ ("w1，w2，…，wm}，wi∈RdWherein d ═ dw + dp × 2.
7. The method of claim 5, wherein the bi-directional LSTM layer has its unidirectional LSTM expressed as follows:
itto the input gate, ftTo forget the door, ctIn a cellular state, otTo the output gate, htFor the hidden vector, Wx，Wh，WeIn order to be the weight, the weight is,
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+ittanh(Wxcxt+Whcht-1+bc)
ot＝σ(Wxoxt+Whoht-1+Wcoct+bo)
ht＝ottanh(ct)
vector h of bidirectional LSTMtOutput from the forward network
Figure FDA0002337580080000021
And reverse output
Figure FDA0002337580080000022
Are obtained together.
8. The method of claim 5, wherein u is the attention level of the word leveliRepresenting the score of the correlation of each word in a sentence, r representing a random query vector, hiIs htEach term of (2), namely each word hidden vector, has the following specific relationship:
ui＝hi·r
αiis the weight obtained by the word-level attention mechanism, and the calculation formula is as follows:
Figure FDA0002337580080000023
s is a vector representation of a sentence, and the calculation formula is as follows:
Figure FDA0002337580080000024
9. the method of claim 5, wherein the sentence-level attention layer is configured to add entity description information;
in the sentence-level attention layer, eiRepresenting an input sentence siRelation to prediction rkThe calculation formula of the matching degree is as follows:
Figure FDA0002337580080000031
αiis the weight obtained by the sentence level attention mechanism, and the calculation formula is as follows:
Figure FDA0002337580080000032
b is a vector representation of a packet, which is equal to the sum of the weights of all sentences, and the calculation formula is as follows:
Figure FDA0002337580080000033
the obtained packet represents description information of the entity on the splice, i.e. a category information vector of the entity, and is represented as:
Figure FDA0002337580080000034
CN201911362557.9A 2019-12-26 2019-12-26 Multi-character relation extraction method based on multi-level attention mechanism Active CN111125367B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911362557.9A CN111125367B (en) 2019-12-26 2019-12-26 Multi-character relation extraction method based on multi-level attention mechanism

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911362557.9A CN111125367B (en) 2019-12-26 2019-12-26 Multi-character relation extraction method based on multi-level attention mechanism

Publications (2)

Publication Number Publication Date
CN111125367A true CN111125367A (en) 2020-05-08
CN111125367B CN111125367B (en) 2023-05-23

Family

ID=70502727

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911362557.9A Active CN111125367B (en) 2019-12-26 2019-12-26 Multi-character relation extraction method based on multi-level attention mechanism

Country Status (1)

Country Link
CN (1) CN111125367B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111651606A (en) * 2020-06-05 2020-09-11 深圳市慧择时代科技有限公司 Text processing method and device and electronic equipment
CN112101009A (en) * 2020-09-23 2020-12-18 中国农业大学 Knowledge graph-based method for judging similarity of people relationship frame of dream of Red mansions
CN112560490A (en) * 2020-12-08 2021-03-26 吉林大学 Knowledge graph relation extraction method and device, electronic equipment and storage medium
CN112818683A (en) * 2021-01-26 2021-05-18 山西三友和智慧信息技术股份有限公司 Chinese character relationship extraction method based on trigger word rule and Attention-BilSTM
CN112926325A (en) * 2021-02-14 2021-06-08 北京工业大学 Chinese character relation extraction construction method based on BERT neural network
CN113128229A (en) * 2021-04-14 2021-07-16 河海大学 Chinese entity relation joint extraction method
CN113919350A (en) * 2021-09-22 2022-01-11 上海明略人工智能(集团)有限公司 Entity identification method, system, electronic equipment and storage medium
CN117057345A (en) * 2023-10-11 2023-11-14 腾讯科技（深圳）有限公司 Role relation acquisition method and related products

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109408812A (en) * 2018-09-30 2019-03-01 北京工业大学 A method of the sequence labelling joint based on attention mechanism extracts entity relationship
CN110502749A (en) * 2019-08-02 2019-11-26 中国电子科技集团公司第二十八研究所 A kind of text Relation extraction method based on the double-deck attention mechanism Yu two-way GRU

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109408812A (en) * 2018-09-30 2019-03-01 北京工业大学 A method of the sequence labelling joint based on attention mechanism extracts entity relationship
CN110502749A (en) * 2019-08-02 2019-11-26 中国电子科技集团公司第二十八研究所 A kind of text Relation extraction method based on the double-deck attention mechanism Yu two-way GRU

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111651606A (en) * 2020-06-05 2020-09-11 深圳市慧择时代科技有限公司 Text processing method and device and electronic equipment
CN111651606B (en) * 2020-06-05 2024-03-01 深圳市慧择时代科技有限公司 Text processing method and device and electronic equipment
CN112101009A (en) * 2020-09-23 2020-12-18 中国农业大学 Knowledge graph-based method for judging similarity of people relationship frame of dream of Red mansions
CN112101009B (en) * 2020-09-23 2024-03-26 中国农业大学 Method for judging similarity of red-building dream character relationship frames based on knowledge graph
CN112560490A (en) * 2020-12-08 2021-03-26 吉林大学 Knowledge graph relation extraction method and device, electronic equipment and storage medium
CN112818683A (en) * 2021-01-26 2021-05-18 山西三友和智慧信息技术股份有限公司 Chinese character relationship extraction method based on trigger word rule and Attention-BilSTM
CN112926325A (en) * 2021-02-14 2021-06-08 北京工业大学 Chinese character relation extraction construction method based on BERT neural network
CN113128229A (en) * 2021-04-14 2021-07-16 河海大学 Chinese entity relation joint extraction method
CN113128229B (en) * 2021-04-14 2023-07-18 河海大学 Chinese entity relation joint extraction method
CN113919350A (en) * 2021-09-22 2022-01-11 上海明略人工智能(集团)有限公司 Entity identification method, system, electronic equipment and storage medium
CN117057345A (en) * 2023-10-11 2023-11-14 腾讯科技（深圳）有限公司 Role relation acquisition method and related products
CN117057345B (en) * 2023-10-11 2024-01-30 腾讯科技（深圳）有限公司 Role relation acquisition method and related products

Also Published As

Publication number Publication date
CN111125367B (en) 2023-05-23

Similar Documents

Publication Publication Date Title
US11631007B2 (en) Method and device for text-enhanced knowledge graph joint representation learning
CN111125367B (en) Multi-character relation extraction method based on multi-level attention mechanism
CN108733792B (en) Entity relation extraction method
CN110502749B (en) Text relation extraction method based on double-layer attention mechanism and bidirectional GRU
CN110119765B (en) Keyword extraction method based on Seq2Seq framework
CN108984526B (en) Document theme vector extraction method based on deep learning
CN107291795B (en) Text classification method combining dynamic word embedding and part-of-speech tagging
CN107085581B (en) Short text classification method and device
CN110765775B (en) Self-adaptive method for named entity recognition field fusing semantics and label differences
CN108009148B (en) Text emotion classification representation method based on deep learning
CN111401061A (en) Method for identifying news opinion involved in case based on BERT and Bi L STM-Attention
CN106569998A (en) Text named entity recognition method based on Bi-LSTM, CNN and CRF
CN110969020A (en) CNN and attention mechanism-based Chinese named entity identification method, system and medium
CN112487807A (en) Text relation extraction method based on expansion gate convolution neural network
CN111241816A (en) Automatic news headline generation method
CN113704546A (en) Video natural language text retrieval method based on space time sequence characteristics
CN112069831A (en) Unreal information detection method based on BERT model and enhanced hybrid neural network
CN111145914B (en) Method and device for determining text entity of lung cancer clinical disease seed bank
CN110276396B (en) Image description generation method based on object saliency and cross-modal fusion features
CN114428850B (en) Text retrieval matching method and system
US20220156489A1 (en) Machine learning techniques for identifying logical sections in unstructured data
CN116932661A (en) Event knowledge graph construction method oriented to network security
CN111881256A (en) Text entity relation extraction method and device and computer readable storage medium equipment
CN113239663A (en) Multi-meaning word Chinese entity relation identification method based on Hopkinson
CN114757184B (en) Method and system for realizing knowledge question and answer in aviation field

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