CN113312487A - Knowledge representation learning method facing legal text based on TransE model - Google Patents
Knowledge representation learning method facing legal text based on TransE model Download PDFInfo
- Publication number
- CN113312487A CN113312487A CN202110058262.3A CN202110058262A CN113312487A CN 113312487 A CN113312487 A CN 113312487A CN 202110058262 A CN202110058262 A CN 202110058262A CN 113312487 A CN113312487 A CN 113312487A
- Authority
- CN
- China
- Prior art keywords
- training
- industry
- text
- learning
- model
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 108
- 239000013598 vector Substances 0.000 claims abstract description 50
- 230000006870 function Effects 0.000 claims abstract description 29
- 239000004576 sand Substances 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 230000000873 masking effect Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- XOAAWQZATWQOTB-UHFFFAOYSA-N taurine Chemical compound NCCS(O)(=O)=O XOAAWQZATWQOTB-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 229960003080 taurine Drugs 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/242—Dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/18—Legal services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Tourism & Hospitality (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Technology Law (AREA)
- Animal Behavior & Ethology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a knowledge representation learning method facing legal texts based on a TransE model, which comprises the following steps: s1: acquiring a legal industry training text by using a mask language model; s2: dividing entities according to the obtained legal industry training texts, extracting corresponding relations among the entities, and storing the defined data to a graph database in a triple form; s3: matching the industry word entities in the training text by using the graph database, and defining a target function of a knowledge representation learning TransE model; and training the model by fusing the entity vector in the training text and the structural information in the graph database, and learning the expression of the entity vector and the relationship vector. The invention solves the problem that the traditional knowledge representation learning method only utilizes structural information and does not utilize a plurality of additional information, so that the knowledge representation fused with text information can better represent the complex relation in a knowledge base.
Description
Technical Field
The invention relates to the field of legal knowledge maps, in particular to a knowledge representation learning method facing legal texts based on a TransE model.
Background
The method is limited by a plurality of defects of the current deep learning technology, only the work of regular data and simpler task can be solved, and in some situations with complex structures, the deep learning technology is difficult to play, and still needs to be distinguished by human experience.
Under the condition, the rise of the knowledge graph technology provides a very convenient and efficient feasible scheme. The knowledge graph stores an event as triple data of an entity-relation-entity, so that the limitation of the traditional database is broken through, and the problem of searching related data is greatly simplified.
The knowledge base may refer to a database in which related data is stored in order. In general, a knowledge base may be represented in a network form, with nodes representing entities and edges representing relationships between entities. In a network-form representation, it is usually necessary to design a specialized knowledge-graph computation storage and to utilize a knowledge base. The design of the knowledge graph not only has the defects of time and labor waste, but also is troubled by the problem of data sparsity. Therefore, a representation learning technique typified by deep learning has attracted attention. Representation learning aims at identifying semantic information of a study object as a dense low-dimensional real-valued vector.
A translation-based model is a typical knowledge representation method, and the model regards relationships as translation operations between entities, i.e., a relationship vector can be represented as a difference between a tail entity vector and a head entity vector. When the relation between the entities is missing, the relation vector can be calculated through the difference of the entity vectors, and the relation corresponding to the relation vector is found out to complete the relation. The model has extremely high accuracy in the knowledge base completion experiment. However, most of the existing translation-based models only use structural information in the knowledge base, and ignore additional information such as relationship path information, type information, entity description information and the like in the knowledge base.
In the field of NLP (Natural Language Processing), pre-training Language models all show excellent effects on multiple NLP tasks. In addition, the pre-trained language model also performs well on tasks that need real-world description and knowledge reasoning, such as many reading understanding tasks, information extraction tasks, and the like, which shows that the pre-trained language model has better knowledge acquisition capability and can be used for better learning knowledge representation.
Therefore, the knowledge base is modeled by combining the text information with the powerful knowledge acquisition capability and the context analysis capability of the pre-training language model, the problem that only structured information is used and multiple kinds of additional information are not used in the traditional knowledge representation learning method can be solved, and the complex relation in the knowledge base can be better represented by the knowledge representation obtained by combining the text information.
Therefore, there is a need to provide a knowledge representation learning method facing legal text based on the TransE model to solve the above problems.
Disclosure of Invention
The invention aims to provide a knowledge representation learning method facing legal texts based on a TransE model, which comprehensively considers the structural information and the text description information of knowledge and improves the accuracy of knowledge representation.
In order to achieve the purpose, the invention provides the following technical scheme: a knowledge representation learning method facing legal text based on a TransE model comprises the following steps: s1: acquiring a legal industry training text by using a mask language model; s2: dividing entities according to the obtained legal industry training texts, extracting corresponding relations among the entities, and storing the defined data to a graph database in a triple form; s3: matching the industry word entities in the training text by using the graph database, and defining a target function of a knowledge representation learning TransE model; and training the model by fusing the entity vector in the training text and the structural information in the graph database, and learning the expression of the entity vector and the relationship vector.
The step of S3 includes the following steps, (1) utilizing hsAnd tsRespectively representing entity vectors in a graph database, and establishing representations of a head entity and a tail entity from the perspective of structural information, wherein the representations are the same as those of a TransE model; (2) matching the industry word entities in the training text by using a graph database to obtain an entity vector representation h based on the training textwAnd twBuilding representations of head entities and tail entities from the perspective of the training text; (3) forming an energy function of the optimized knowledge representation learning model:
E(h,r,t)=||hs+r-ts||+||hw+r-tw||+||hs+r-tw||+ ||hw+r-ts||
wherein, the first part is an energy function based on the structural representation, the second part is an energy function based on the text information, and the third part and the fourth part are energy functions based on the fusion of the structural information and the text information; (4) obtaining embedded representations of entities and relationships through an objective function of a knowledge representation learning model that exhibits characteristics of the entities and relationships in a knowledge graph, the objective function being:
in the formula, gamma is an edge hyper-parameter, T is a training set, and T' is a negative sample set of T; (5) h obtained by pre-training a language modelwAnd twRandomly initialized hs、r、tsAnd as the initial input of a knowledge representation learning model, optimizing an objective function by using a random gradient descent method according to the training mode of a TransE model, training and solving the model, and learning the representation of an entity vector and a relation vector.
Compared with the prior art, the knowledge representation learning method facing the legal text based on the TransE model comprehensively considers the structural information and the text description information of knowledge, improves the accuracy of knowledge representation, and aims to more fully represent the characteristics of entities and relations in a knowledge graph so as to obtain more efficient embedded representation of the entities and relations.
The method utilizes the strong knowledge acquisition capability and the context analysis capability of the pre-training language model, integrates text information to model the knowledge base, and solves the problem that the traditional knowledge representation learning method only utilizes structural information and does not utilize a plurality of additional information, so that the knowledge representation integrated with the text information can better represent the complex relation in the knowledge base.
Drawings
FIG. 1 is a flow chart diagram of a knowledge representation learning method facing legal text based on a TransE model;
FIG. 2 is a schematic flow chart of training of knowledge representation learning in the knowledge representation learning method for legal text according to the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the present invention provides a knowledge representation learning method for legal text based on a TransE model, which includes the following steps:
s1: acquiring a legal industry training text by using a mask language model; specifically, a mask language model is used for masking industry words in a training text to obtain a mask training text; inputting the mask training text into a pre-training language model, and learning to obtain knowledge representation of an industry word in a legal text;
the customizing step of the mask language model comprises the following steps:
collecting the industry corpora;
performing word segmentation on the industry linguistic data to obtain a word set;
counting the distribution of the words in the word set to obtain a distribution result;
and selecting words from the word set based on the distribution result, and generating an industry dictionary as the mask language model.
The acquiring of the industry training text comprises:
collecting an industrial question and answer corpus;
taking the industry question-answer corpus as a formal industry training text;
and scattering question sentences and answer sentences in the industry question-answer corpus to generate a negative example industry training text.
Wherein the industry is a legal industry, and the obtaining an industry training text comprises:
collecting legal decision corpus; and deleting case routing information in the legal decision book corpus to generate a legal industry training text.
Wherein, the acquiring of the industry training text further comprises:
inserting a first preset character into the head of the legal industry training text, dividing the legal industry training text according to a fixed character length, and inserting a second preset character into the tail of each divided part;
and inputting the mask training text into a pre-training language model, and learning to obtain knowledge representation of the industry words in the industry training text and case relations to which the legal industry training text belongs.
S2: dividing entities according to the obtained legal industry training texts, extracting corresponding relations among the entities, and storing the defined data to a database in a triple form;
triples, entities in a knowledge graph refer to words having a concrete or abstract meaning, relationships refer to associations between different entities, which are typically stored in the form of triples (head entity h, relationship between head and tail entities r, tail entity t);
the knowledge representation of the industry words refers to that the mask language model is used for masking the industry words in the legal industry texts to obtain mask training texts, and then the mask training texts are input into the pre-training language model to obtain the knowledge representation of entity (industry word) vectors in the legal texts.
Specifically, the division entity includes five categories of entities, namely, a person, a case (event), an article, a place and an organization.
The graph database is used for storing various relational graphs, each node in the graph represents an entity, and edges between the nodes represent relations, so that defined data can be stored into the graph database in a node-edge-node mode. The step of constructing the graph database comprises:
acquiring all relevant data of the broken cases, including case time, case place, case article, case personnel and all personnel data related to the case article, dividing the data into five types of entities of personnel, cases, articles, places and mechanisms, and extracting the relationship among the five types of entities;
and storing the extracted events in a form of entity, relation and entity as a format of a triple, and marking the symbol as (h, r, t), wherein h represents a main entity of the event, r represents the relation, and t represents a guest entity of the event.
S3: matching the industry word entities in the training text by using a graph database, and defining a target function of a knowledge representation learning TransE model; and training the model by fusing the entity vector in the training text and the structural information in the graph database, and learning the expression of the entity vector and the relationship vector.
Fig. 2 is a schematic flow chart of training of knowledge representation learning in the knowledge representation learning method for legal text according to the present invention. The training of knowledge representation learning is to adopt an optimized TransE model to fuse the training text and graph database structure information of the legal industry, and the method comprises the following steps:
(1) by using hsAnd tsRespectively representing entity vectors in a graph database, and establishing representations of a head entity and a tail entity from the perspective of structural information, wherein the representations are the same as those of a TransE model;
(2) matching the industry word entities in the training text by using a graph database to obtain an entity vector representation h based on the training textwAnd twBuilding representations of head entities and tail entities from the perspective of the training text;
the method for obtaining the entity vector representation by matching the pre-training language model with the graph database comprises the following steps:
and (2.1) acquiring a legal industry training text. Typically, industry words will be present in industry training text. An industry word may be a word that is unique to the industry, i.e., a word that has only a specific meaning in the industry. For example, "taurine particles" are an industry word for the medical industry.
(2.2) masking the industry words in the legal industry training text by using a mask language model to obtain a mask training text;
(2.3) inputting the mask training text into a pre-training language model, and learning to obtain knowledge representation of the industry words in the industry training text;
(2.4) matching the entity in the graph database with the legal training text to further obtain the entity vector representation, namely h, learned by the pre-training language modelwAnd tw。
(3) Forming an energy function of the optimized knowledge representation learning model:
E(h,r,t)=||hs+r-ts||+||hw+r-tw||+||hs+r-tw||+ ||hw+r-ts||
in the formula, the first part is an energy function based on the structural representation, the second part is an energy function based on the text information, and the third part and the fourth part are energy functions based on the fusion of the structural information and the text information. The energy function maps the two types of entity representations to the same vector space containing all four energy function relational representations, which are facilitated by each other;
text information can be well fused through the learning model; and further preparing expression entities and relationship vectors.
The mask language model can be used for mask industry words and is customized by utilizing a large-scale industry dictionary, and the large-scale industry dictionary can be mined from mass industry linguistic data by utilizing a data mining technology. The industry dictionary mining method comprises the following steps:
firstly, collecting industry linguistic data;
secondly, performing word segmentation on the industry linguistic data to obtain a word set;
then, the distribution of the words in the word set is counted to obtain a distribution result. Here, the frequency of each word appearing in the line corpus is counted;
and finally, selecting words from the word set based on the distribution result to generate an industry dictionary. Wherein, the industry dictionary can be used as a mask language model. Here, the word with high frequency of occurrence is selected firstly, and then the universal word is deleted manually, so that the industry dictionary can be obtained.
Specifically, an industry word in the legal industry training text is recognized by using a Mask Language Model (Mask Language Model), and then the industry word is shielded.
(4) The embedded representation of the entity and the relation is obtained through the objective function of the knowledge representation learning model, the characteristics of the entity and the relation in the knowledge map are more fully represented, and the method has the advantages that: thereby obtaining a more efficient embedded representation of entities and relationships. The objective function is as follows:
in the formula, gamma is an edge hyper-parameter, T is a training set, and T' is a negative sample set of T;
(5) h obtained by pre-training a language modelwAnd twRandomly initialized hs、r、 tsAnd as initial input of a knowledge representation learning model, optimizing an objective function by using a random gradient descent method according to a training mode of a TransE model, training and solving the model, and learning representation of entity vectors and relationship vectors.
The specific training method of the knowledge representation learning model comprises the following steps:
(5.1) determining a training set, a hyper-parameter gamma, a learning rate lambda, an embedded dimension k and a pre-trained text entity vector hwAnd tw;
(5.2) initializing relationship vectors and entity vectors in the graph database, for each dimension of each vector, inA value is randomly selected, k is the dimension of the low-dimensional vector, and normalization is carried out after all vectors are initialized;
(5.3) entering a circulation: by adopting minipatch, training of one batch of data can accelerate the training speed, each batch of data is subjected to negative sampling (one entity of a triplet in a training set is randomly replaced), T _ batch is initially an empty list, and then a list consisting of tuple pairs (original triplets, smashed triplets) is added to the empty list:
T_batch=[([h,r,t],[h',r,t']),……]
and training after the T _ batch is obtained, adjusting parameters by adopting gradient descent, and optimizing the target function until the iteration number reaches the preset maximum iteration number, so as to realize the learning of knowledge expression.
The invention collaboratively learns the structural information and the training text description information in the knowledge base, and can obtain the following beneficial effects by the technical scheme of the knowledge representation learning method facing legal texts, wherein the technical scheme is conceived by the knowledge representation learning method facing the legal texts:
the knowledge representation learning algorithm is used for embedding entities and relations in the triples, training and learning are carried out on the basis of the established knowledge graph by combining with the text description information, and the knowledge representation learning algorithm has important significance for reasoning work.
The calculation efficiency is remarkably improved, the semantic and reasoning relation among the entities is calculated by adopting a graph algorithm for the knowledge graph in the prior art, the calculation complexity is high, the expandability is poor, the distributed representation obtained by representation learning can be realized, and the operations such as semantic similarity calculation and the like can be realized efficiently.
The problem of data sparsity is effectively relieved, entities are projected to a uniform low-dimensional space through representation and learning, each entity corresponds to one dense vector, and therefore semantic similarity between any two entities can be measured.
Heterogeneous information fusion is realized, entities from different sources are projected into the same semantic space through a representation learning model, a uniform representation space is established, semantic similarity correlation calculation among heterogeneous entities is realized, and information fusion of multiple knowledge bases is realized.
Claims (7)
1. A knowledge representation learning method facing legal text based on TransE model is characterized by comprising the following steps:
s1: acquiring a legal industry training text by using a mask language model;
s2: dividing entities according to the obtained legal industry training texts, extracting corresponding relations among the entities, and storing the defined data to a graph database in a triple form;
s3: matching the industry word entities in the training text by using the graph database, and defining a target function of a knowledge representation learning TransE model; and training the model by fusing the entity vector in the training text and the structural information in the graph database, and learning the expression of the entity vector and the relationship vector.
2. The method of learning legal text-oriented knowledge representation based on TransE model according to claim 1, wherein the step S3 includes the steps of,
(1) by using hsAnd tsRespectively representing entity vectors in a graph database, and establishing representations of a head entity and a tail entity from the perspective of structural information, wherein the representations are the same as those of a TransE model;
(2) matching the industry word entities in the training text by using a graph database to obtain an entity vector representation h based on the training textwAnd twBuilding representations of head entities and tail entities from the perspective of the training text;
(3) forming an energy function of the optimized knowledge representation learning model:
E(h,r,t)=‖hs+r-ts‖+‖hw+r-tw‖+‖hs+r-tw‖+‖hw+r-ts‖
wherein, the first part is an energy function based on the structural representation, the second part is an energy function based on the text information, and the third part and the fourth part are energy functions based on the fusion of the structural information and the text information;
(4) obtaining embedded representations of entities and relationships through an objective function of a knowledge representation learning model that exhibits characteristics of the entities and relationships in a knowledge graph, the objective function being:
in the formula, gamma is an edge hyper-parameter, T is a training set, and T' is a negative sample set of T;
(5) h obtained by pre-training a language modelwAnd twRandomly initialized hs、r、tsAnd as the initial input of a knowledge representation learning model, optimizing an objective function by using a random gradient descent method according to the training mode of a TransE model, training and solving the model, and learning the representation of an entity vector and a relation vector.
3. The method of learning legal text-oriented knowledge representation based on TransE model according to claim 2, wherein the method of matching a pre-trained language model with a graph database to obtain an entity vector representation comprises the steps of:
obtaining a legal industry training text;
masking the industry words in the legal industry training text by using a mask language model to obtain a mask training text;
inputting the mask training text into a pre-training language model, and learning to obtain knowledge representation of industry words in the industry training text;
matching the entity in the graph database with the legal training text to obtain the entity vector representation, namely h, learned by the pre-training language modelwAnd tw。
4. The method of claim 3, wherein the specific training method of the knowledge representation learning model comprises:
determining a training set, a hyper-parameter gamma, a learning rate lambda, an embedded dimension k and a pre-trained text entity vector hwAnd tw;
Initializing relationship vectors and entity vectors in a graph database for each dimension of each vectorTaking a value at random, k being the dimension of the low-dimensional vector, for all directionsNormalization is performed after quantity initialization;
entering a circulation: with minipatch, training of a batch of data can be accelerated, negative sampling is performed on each batch of data, T _ batch is initially an empty list, and then a list consisting of tuple groups is added thereto:
T_batch=[([h,r,t],[h',r,t']),([],[]),……];
and training after the T _ batch is obtained, adjusting parameters by adopting gradient descent, and optimizing the target function until the iteration number reaches the preset maximum iteration number, so as to realize the learning of knowledge expression.
5. The method of learning legal text-oriented knowledge representation based on TransE model according to claim 1, wherein the step of customizing the mask language model comprises:
collecting the industry corpora;
performing word segmentation on the industry linguistic data to obtain a word set;
counting the distribution of the words in the word set to obtain a distribution result;
and selecting words from the word set based on the distribution result, and generating an industry dictionary as the mask language model.
6. The method of learning legal text-oriented knowledge representation based on TransE model according to claim 1, wherein the obtaining of industry training text comprises:
collecting an industrial question and answer corpus;
taking the industry question-answer corpus as a formal industry training text;
and scattering question sentences and answer sentences in the industry question-answer corpus to generate a negative example industry training text.
7. The method of learning legal text-oriented knowledge representation based on TransE model according to claim 6, wherein the obtaining of industry training text further comprises:
inserting first preset characters into the heads of the positive example industry training texts and the negative example industry training texts, and inserting second preset characters into the ends of the question sentences and the answer sentences;
and inputting the mask training text into a pre-training language model, and learning to obtain the knowledge representation of the industry words in the industry training text and the positive and negative case prediction values of the industry training text.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110058262.3A CN113312487A (en) | 2021-01-16 | 2021-01-16 | Knowledge representation learning method facing legal text based on TransE model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110058262.3A CN113312487A (en) | 2021-01-16 | 2021-01-16 | Knowledge representation learning method facing legal text based on TransE model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113312487A true CN113312487A (en) | 2021-08-27 |
Family
ID=77370612
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110058262.3A Pending CN113312487A (en) | 2021-01-16 | 2021-01-16 | Knowledge representation learning method facing legal text based on TransE model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113312487A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180341863A1 (en) * | 2017-05-27 | 2018-11-29 | Ricoh Company, Ltd. | Knowledge graph processing method and device |
CN109033129A (en) * | 2018-06-04 | 2018-12-18 | 桂林电子科技大学 | Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method |
CN110334219A (en) * | 2019-07-12 | 2019-10-15 | 电子科技大学 | The knowledge mapping for incorporating text semantic feature based on attention mechanism indicates learning method |
CN111008186A (en) * | 2019-06-11 | 2020-04-14 | 中央民族大学 | Expression method of Tibetan knowledge base |
US20200218988A1 (en) * | 2019-01-08 | 2020-07-09 | International Business Machines Corporation | Generating free text representing semantic relationships between linked entities in a knowledge graph |
CN111581395A (en) * | 2020-05-06 | 2020-08-25 | 西安交通大学 | Model fusion triple representation learning system and method based on deep learning |
CN111680145A (en) * | 2020-06-10 | 2020-09-18 | 北京百度网讯科技有限公司 | Knowledge representation learning method, device, equipment and storage medium |
CN111723021A (en) * | 2020-07-23 | 2020-09-29 | 哈尔滨工业大学 | Defect report automatic allocation method based on knowledge base and representation learning |
-
2021
- 2021-01-16 CN CN202110058262.3A patent/CN113312487A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180341863A1 (en) * | 2017-05-27 | 2018-11-29 | Ricoh Company, Ltd. | Knowledge graph processing method and device |
CN109033129A (en) * | 2018-06-04 | 2018-12-18 | 桂林电子科技大学 | Multi-source Information Fusion knowledge mapping based on adaptive weighting indicates learning method |
US20200218988A1 (en) * | 2019-01-08 | 2020-07-09 | International Business Machines Corporation | Generating free text representing semantic relationships between linked entities in a knowledge graph |
CN111008186A (en) * | 2019-06-11 | 2020-04-14 | 中央民族大学 | Expression method of Tibetan knowledge base |
CN110334219A (en) * | 2019-07-12 | 2019-10-15 | 电子科技大学 | The knowledge mapping for incorporating text semantic feature based on attention mechanism indicates learning method |
CN111581395A (en) * | 2020-05-06 | 2020-08-25 | 西安交通大学 | Model fusion triple representation learning system and method based on deep learning |
CN111680145A (en) * | 2020-06-10 | 2020-09-18 | 北京百度网讯科技有限公司 | Knowledge representation learning method, device, equipment and storage medium |
CN111723021A (en) * | 2020-07-23 | 2020-09-29 | 哈尔滨工业大学 | Defect report automatic allocation method based on knowledge base and representation learning |
Non-Patent Citations (1)
Title |
---|
肖宝,韩开旭: "基于主题模型和词嵌入的知识表示学习的研究", 《北部湾大学学报》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107766324B (en) | Text consistency analysis method based on deep neural network | |
CN109902159A (en) | A kind of intelligent O&M statement similarity matching process based on natural language processing | |
CN111078836B (en) | Machine reading understanding method, system and device based on external knowledge enhancement | |
Zhang et al. | DeepDive: Declarative knowledge base construction | |
Mansinghka et al. | Structured priors for structure learning | |
CN112883738A (en) | Medical entity relation extraction method based on neural network and self-attention mechanism | |
CN108038205B (en) | Viewpoint analysis prototype system for Chinese microblogs | |
CN111274790B (en) | Chapter-level event embedding method and device based on syntactic dependency graph | |
CN111966812B (en) | Automatic question answering method based on dynamic word vector and storage medium | |
CN111767325B (en) | Multi-source data deep fusion method based on deep learning | |
CN112988917A (en) | Entity alignment method based on multiple entity contexts | |
CN111581368A (en) | Intelligent expert recommendation-oriented user image drawing method based on convolutional neural network | |
Panda | Developing an efficient text pre-processing method with sparse generative Naive Bayes for text mining | |
CN113988075A (en) | Network security field text data entity relation extraction method based on multi-task learning | |
Sadr et al. | Unified topic-based semantic models: A study in computing the semantic relatedness of geographic terms | |
CN115130538A (en) | Training method of text classification model, text processing method, equipment and medium | |
Panicker et al. | Image caption generator | |
CN114373554A (en) | Drug interaction relation extraction method using drug knowledge and syntactic dependency relation | |
CN116720519B (en) | Seedling medicine named entity identification method | |
Kusuma et al. | Automatic question generation with classification based on mind map | |
CN116049376B (en) | Method, device and system for retrieving and replying information and creating knowledge | |
Pujianto et al. | Text Difficulty Classification Based on Lexile Levels Using K-Means Clustering and Multinomial Naive Bayes | |
CN113312487A (en) | Knowledge representation learning method facing legal text based on TransE model | |
Huang et al. | An effective method for constructing knowledge graph of online course | |
CN111581339B (en) | Method for extracting gene events of biomedical literature based on tree-shaped LSTM |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210827 |
|
WD01 | Invention patent application deemed withdrawn after publication |