CN112559765B - Semantic integration method for multi-source heterogeneous database - Google Patents
Semantic integration method for multi-source heterogeneous database Download PDFInfo
- Publication number
- CN112559765B CN112559765B CN202011440234.XA CN202011440234A CN112559765B CN 112559765 B CN112559765 B CN 112559765B CN 202011440234 A CN202011440234 A CN 202011440234A CN 112559765 B CN112559765 B CN 112559765B
- Authority
- CN
- China
- Prior art keywords
- entity
- matching
- candidate
- aligned
- candidate 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.)
- Active
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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a multi-source heterogeneous database semantic integration method, which comprises the following steps: (1) extraction entity: extracting field related entities from unstructured text based on the entity extraction model and identifying corresponding categories; (2) concept matching: matching the ontology concept in the knowledge graph according to the corresponding category to obtain a candidate entity set of the same category; (3) neighborhood matching: obtaining an aligned entity graph representation according to the context information of the related entity, and obtaining a candidate entity graph representation according to the domain relation of the candidate entity set in the knowledge graph; (4) and (3) comparison decision: and comparing and deciding the aligned entity graph representation and the candidate entity graph representation to obtain the best matching candidate entity arrangement as a matching result. The invention combines the deep reinforcement learning technology with the multi-source heterogeneous database semantic integration, establishes the semantic mapping relation between knowledge under different forms, and can better support related applications such as semantic retrieval, intelligent question-answering and the like based on the semantic integration.
Description
Technical Field
The invention relates to a semantic integration method for a multi-source heterogeneous database.
Background
With the development of the information society, the fragmentation problem of the multi-source heterogeneous database is more and more prominent. In the big data age, the current information resource utilization mode is changing from information management by means of homologous structured data to information integration management by multi-source heterogeneous resource sharing. However, the current integration effect based on heterogeneous data has not been able to accommodate increasingly complex application requirements. How to effectively utilize the structured and unstructured data and combine the independent and distributed databases has important significance for realizing information sharing and improving the data utilization value. How to realize the semantic fusion of the unstructured text database and the structured knowledge graph is an outstanding problem of the semantic integration of the multi-source heterogeneous database.
Disclosure of Invention
In order to solve the technical problems, the invention provides a multi-source heterogeneous database semantic integration method, which combines a deep reinforcement learning technology with multi-source heterogeneous database semantic integration and can effectively establish semantic mapping relations among knowledge under different forms.
The invention is realized by the following technical scheme.
The invention provides a multi-source heterogeneous database semantic integration method, which comprises the following steps:
(1) extraction entity: extracting field related entities from unstructured text based on the entity extraction model and identifying corresponding categories;
(2) concept matching: matching the ontology concept in the knowledge graph according to the corresponding category to obtain a candidate entity set of the same category;
(3) neighborhood matching: obtaining an aligned entity graph representation according to the context information of the related entity, and obtaining a candidate entity graph representation according to the domain relation of the candidate entity set in the knowledge graph;
(4) and (3) comparison decision: and comparing and deciding the aligned entity graph representation and the candidate entity graph representation to obtain the best matching candidate entity arrangement as a matching result.
The method also comprises the following steps:
(5) feedback optimization: and optimizing parameters of the entity extraction model by using evaluation feedback of the matching result.
The step (1) comprises the following steps:
s11: marking the unstructured text according to a label system, and constructing a data set aiming at the extraction entity task;
s12: based on the sequence labeling model, extracting related entities of the unstructured text, and identifying the category corresponding to the related entities.
The sequence labeling model is constructed by combining a pre-training language model BERT with a conditional random field CRF.
The step (3) of neighborhood matching comprises the following steps:
s31: carrying out syntactic dependency analysis on the context of the unstructured text of the related entity to obtain an aligned entity diagram representation to be aligned;
s32: and according to the candidate entity set, obtaining a candidate entity graph representation of each candidate entity through the position of each candidate entity in the candidate entity set in the knowledge graph and the neighborhood structural characteristics of the candidate entity.
The step (4) of comparing decisions comprises the following steps:
s41: using the alignment entity diagram representation and the candidate entity diagram representation as input, and using a strategy network to output a matching result of the entity to be aligned and the candidate entity;
s42: and returning N candidate entities with highest matching degree in the matching result and a matching degree sorting list.
The strategy network is a twin-map neural network model.
The step (5) of feedback optimization comprises the following steps:
s51: collecting evaluation feedback of the matching result of each entity to be aligned;
s52: according to the matching degree sequencing list predicted by the strategy network for each entity to be aligned, constructing a reward function aiming at the strategy network by combining evaluation feedback;
s53: recording historical input and output of a strategy network and corresponding rewards as a sample training set;
s54: and when the number of samples in the sample training set is greater than M, starting a training process of the strategy network.
The invention has the beneficial effects that: the deep reinforcement learning technology is combined with the multi-source heterogeneous database semantic integration, so that semantic mapping relations among knowledge in different forms are established, and related applications such as semantic retrieval, intelligent question-answering and the like based on the semantic integration can be supported better.
Drawings
Fig. 1 is a schematic flow chart of the present invention.
Detailed Description
The technical solution of the present invention is further described below, but the scope of the claimed invention is not limited to the above.
The multi-source heterogeneous database semantic integration method shown in fig. 1 comprises the following steps:
(1) extraction entity: extracting field related entities from unstructured text based on the entity extraction model and identifying corresponding categories;
(2) concept matching: matching the ontology concept in the knowledge graph according to the corresponding category to obtain a candidate entity set of the same category;
(3) neighborhood matching: obtaining an aligned entity graph representation according to the context information of the related entity, and obtaining a candidate entity graph representation according to the domain relation of the candidate entity set in the knowledge graph;
(4) and (3) comparison decision: and comparing and deciding the aligned entity graph representation and the candidate entity graph representation to obtain the best matching candidate entity arrangement as a matching result.
The method also comprises the following steps:
(5) feedback optimization: and optimizing parameters of the entity extraction model by using evaluation feedback of the matching result.
The step (1) comprises the following steps:
s11: marking the unstructured text according to a label system, and constructing a data set aiming at the extraction entity task;
s12: based on the sequence labeling model, extracting related entities of the unstructured text, and identifying the category corresponding to the related entities.
The sequence labeling model is constructed by combining a pre-training language model BERT with a conditional random field CRF.
Step (3) neighborhood matching, comprising the following steps:
s31: carrying out syntactic dependency analysis on the context of the unstructured text of the related entity to obtain an aligned entity diagram representation to be aligned;
s32: and according to the candidate entity set, obtaining a candidate entity graph representation of each candidate entity through the position of each candidate entity in the candidate entity set in the knowledge graph and the neighborhood structural characteristics thereof.
Step (4) of comparing decisions, comprising the steps of:
s41: using the alignment entity diagram representation and the candidate entity diagram representation as input, and using a strategy network to output a matching result of the entity to be aligned and the candidate entity;
s42: and returning N candidate entities with highest matching degree in the matching result and a matching degree sorting list.
The policy network is a twin-map neural network model.
Step (5) feedback optimization, comprising the following steps:
s51: collecting evaluation feedback of the matching result of each entity to be aligned;
s52: according to the matching degree sequencing list predicted by the strategy network for each entity to be aligned, constructing a reward function aiming at the strategy network by combining evaluation feedback;
s53: recording historical input and output of a strategy network and corresponding rewards as a sample training set;
s54: and when the number of samples in the sample training set is greater than M, starting a training process of the strategy network.
Example 1
By adopting the scheme, the method specifically comprises the following steps:
s1: extracting field related entities from unstructured texts through an entity extraction model based on a deep learning algorithm to obtain start and stop positions of the entities, and identifying categories corresponding to the entities;
s2: matching the identified category of the entity to be aligned with the ontology concept in the knowledge graph to obtain a candidate set with the same category as the entity to be aligned;
s3: obtaining a graph representation of an entity to be aligned according to entity context information of the unstructured text, and obtaining a graph representation of a candidate entity according to a neighborhood relation of nodes in the knowledge graph;
s4: comparing the graph representation of the candidate entities in the candidate set with the graph representation of the entity to be aligned one by one through the deep reinforcement learning model, and thus obtaining candidate entity arrangement which is most matched with the entity to be aligned;
s5: and optimizing parameters of the deep reinforcement learning model by using evaluation feedback of the user on the final matching result, and learning a matching strategy.
Example 2
By adopting the scheme, the method specifically comprises the following steps:
s11: a label system is established, the unstructured text is marked according to the label system, and a data set aiming at entity extraction tasks is established;
s12: and constructing a sequence labeling model of the BERT combined with the conditional random field CRF by utilizing a pre-training language model BERT. Based on the model, completing entity extraction of the residual unstructured text, obtaining start and stop positions of the entity, and identifying a category corresponding to the entity;
wherein: for specific field application, the pre-training language model BERT can firstly make field adaptation in a large-scale non-labeled field-related text corpus; for specific application tasks, task adaptation can be performed in task-related text corpus. To improve the performance of the language model BERT in entity extraction tasks.
S21: and matching the identified category of the entity to be aligned with the ontology concept in the knowledge graph according to the entity extraction result of the unstructured text. And taking the entity set under the matched concept in the knowledge graph as a candidate set of the entity to be aligned.
Wherein: the ontology matching is to build partition indexes for semantic integration tasks of the multi-source heterogeneous database, so that the efficiency of entity linking is improved.
S31: according to the context of the unstructured text of the entity to be aligned, obtaining a graph representation of the entity to be aligned through syntactic dependency analysis of the context;
s32: and according to the candidate set of the entity to be aligned in the S2, for each candidate entity in the candidate set, obtaining the graph representation of the candidate entity through the position of the candidate entity in the knowledge graph and the neighborhood structural characteristics thereof.
The diagram of the entity to be aligned represents the syntactic structure between words in the context of the entity and is used as one of the inputs of the twin-map neural network model; the graph representation of the candidate entity embodies the semantic relationship between the candidate entity and the entity in the domain of the candidate entity in the knowledge graph and is used as one of the inputs of the twin-graph neural network model.
S41: and constructing a twin-map neural network model serving as a strategy network for deep reinforcement learning. The strategy network takes the obtained entity diagram representation to be aligned and the candidate entity diagram representation as input together, and outputs whether the entity to be aligned is matched with the candidate entity;
s42: and aiming at the candidate set of the entity to be aligned, judging the graph representation of each candidate entity in the candidate set and the entity to be aligned one by one through a strategy network, and obtaining the corresponding matching degree. And returning the 10 candidate entities with the highest matching degree and the corresponding matching degree sorting list.
Wherein: the application of the multi-layer twin-map neural network model with the attention mechanism is beneficial to finding and utilizing the map representation of the entity to be aligned and rich map structure information in the map representation of the candidate entity, and is beneficial to predicting the matching degree of the entity to be aligned and the candidate entity by the strategy network.
S51: collecting evaluation feedback of a user on a matching result of each entity to be aligned;
s52: according to the matching degree sequencing list predicted by the strategy network in the S4 for each entity to be aligned, constructing a reward function aiming at the deep reinforcement learning model by combining user evaluation feedback;
s53: recording input, decision output and obtained rewarding values based on a strategy network, and constructing a training set of a deep reinforcement learning model as a sample;
s54: when the collected training samples are larger than a certain number, starting a training program of the strategy network, optimizing parameters of the deep reinforcement learning model, and enabling the model to learn a matching strategy matched with the task.
Wherein: the evaluation feedback of the matching result of each entity to be aligned by the user can be operated as follows: and selecting candidate entities which are semantically equivalent to the entity to be aligned from the matching degree sorting list, or considering that the candidate entities which are semantically equivalent to the entity to be aligned do not exist in the matching degree sorting list. The bonus function may be designed to: if the candidate entity semantically equivalent to the entity to be aligned exists in the matching degree sequencing list, setting the reciprocal of the position of the semantically equivalent candidate entity in the matching degree sequencing list as a reward value; if there is no candidate entity in the matching degree sorting list that is semantically equivalent to the entity to be aligned, the reward value is set to a negative value.
Claims (4)
1. A multi-source heterogeneous database semantic integration method is characterized in that: the method comprises the following steps:
(1) extraction entity: extracting field related entities from unstructured text based on the entity extraction model and identifying corresponding categories;
(2) concept matching: matching the ontology concept in the knowledge graph according to the corresponding category to obtain a candidate entity set of the same category;
(3) neighborhood matching: obtaining an aligned entity graph representation according to the context information of the related entity, and obtaining a candidate entity graph representation according to the domain relation of the candidate entity set in the knowledge graph;
(4) and (3) comparison decision: comparing and deciding the aligned entity graph representation and the candidate entity graph representation to obtain the best matching candidate entity arrangement as a matching result;
the method also comprises the following steps:
(5) feedback optimization: optimizing parameters of the entity extraction model by using evaluation feedback of the matching result;
the step (3) of neighborhood matching comprises the following steps:
s31: carrying out syntactic dependency analysis on the context of the unstructured text of the related entity to obtain an aligned entity diagram representation to be aligned;
s32: according to the candidate entity set, a candidate entity diagram representation of each candidate entity is obtained through the position of each candidate entity in the candidate entity set in the knowledge graph and the neighborhood structural characteristics of the candidate entity;
the step (4) of comparing decisions comprises the following steps:
s41: using the alignment entity diagram representation and the candidate entity diagram representation as input, and using a strategy network to output a matching result of the entity to be aligned and the candidate entity;
s42: returning N candidate entities with highest matching degree in the matching result and a matching degree sorting list;
the step (5) of feedback optimization comprises the following steps:
s51: collecting evaluation feedback of the matching result of each entity to be aligned;
s52: according to the matching degree sequencing list predicted by the strategy network for each entity to be aligned, constructing a reward function aiming at the strategy network by combining evaluation feedback;
s53: recording historical input and output of a strategy network and corresponding rewards as a sample training set;
s54: and when the number of samples in the sample training set is greater than M, starting a training process of the strategy network.
2. The multi-source heterogeneous database semantic integration method according to claim 1, wherein: the step (1) comprises the following steps:
s11: marking the unstructured text according to a label system, and constructing a data set aiming at the extraction entity task;
s12: based on the sequence labeling model, extracting related entities of the unstructured text, and identifying the category corresponding to the related entities.
3. The multi-source heterogeneous database semantic integration method according to claim 2, wherein: the sequence labeling model is constructed by combining a pre-training language model BERT with a conditional random field CRF.
4. The multi-source heterogeneous database semantic integration method according to claim 1, wherein: the strategy network is a twin-map neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011440234.XA CN112559765B (en) | 2020-12-11 | 2020-12-11 | Semantic integration method for multi-source heterogeneous database |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011440234.XA CN112559765B (en) | 2020-12-11 | 2020-12-11 | Semantic integration method for multi-source heterogeneous database |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112559765A CN112559765A (en) | 2021-03-26 |
| CN112559765B true CN112559765B (en) | 2023-06-16 |
Family
ID=75061037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011440234.XA Active CN112559765B (en) | 2020-12-11 | 2020-12-11 | Semantic integration method for multi-source heterogeneous database |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112559765B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113707339B (en) * | 2021-08-02 | 2022-12-09 | 西安交通大学 | Method and system for concept alignment and content inter-translation among multi-source heterogeneous databases |
| CN116150405B (en) * | 2023-04-19 | 2023-06-27 | 中电科大数据研究院有限公司 | Heterogeneous data processing method for multiple scenes |
| CN116821712B (en) * | 2023-08-25 | 2023-12-19 | 中电科大数据研究院有限公司 | Semantic matching method and device for unstructured text and knowledge graph |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145071A (en) * | 2018-08-06 | 2019-01-04 | 中国地质大学（武汉） | A kind of automated construction method and system towards geophysics field knowledge mapping |
| CN109960810A (en) * | 2019-03-28 | 2019-07-02 | 科大讯飞(苏州)科技有限公司 | A kind of entity alignment schemes and device |
| CN110674274A (en) * | 2019-09-23 | 2020-01-10 | 中国农业大学 | Knowledge graph construction method for food safety regulation question-answering system |
| CN111475629A (en) * | 2020-03-31 | 2020-07-31 | 渤海大学 | Knowledge graph construction method and system for math tutoring question-answering system |
| CN111522961A (en) * | 2020-04-09 | 2020-08-11 | 武汉理工大学 | Attention mechanism and entity description based industrial map construction method |
| CN111552820A (en) * | 2020-04-30 | 2020-08-18 | 江河瑞通（北京）技术有限公司 | Water engineering scheduling data processing method and device |
| CN111639190A (en) * | 2020-04-30 | 2020-09-08 | 南京理工大学 | Medical knowledge map construction method |
| CN111737496A (en) * | 2020-06-29 | 2020-10-02 | 东北电力大学 | Power equipment fault knowledge map construction method |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160335544A1 (en) * | 2015-05-12 | 2016-11-17 | Claudia Bretschneider | Method and Apparatus for Generating a Knowledge Data Model |
| US11003661B2 (en) * | 2015-09-04 | 2021-05-11 | Infotech Soft, Inc. | System for rapid ingestion, semantic modeling and semantic querying over computer clusters |
| CN108171184B (en) * | 2018-01-03 | 2020-04-10 | 南京理工大学 | Method for re-identifying pedestrians based on Simese network |
| CN110209832A (en) * | 2018-08-08 | 2019-09-06 | 腾讯科技（北京）有限公司 | Method of discrimination, system and the computer equipment of hyponymy |
| CN111191044A (en) * | 2019-12-25 | 2020-05-22 | 湖北大学 | Knowledge extraction and fusion method based on big data |
| CN111613339B (en) * | 2020-05-15 | 2021-07-09 | 山东大学 | Similar medical record searching method and system based on deep learning |
| CN111814480B (en) * | 2020-07-21 | 2024-04-16 | 华润数字科技有限公司 | Knowledge graph completion method and device, computer equipment and storage medium |
| CN111858940B (en) * | 2020-07-27 | 2023-07-25 | 湘潭大学 | Multi-head attention-based legal case similarity calculation method and system |
-
2020
- 2020-12-11 CN CN202011440234.XA patent/CN112559765B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145071A (en) * | 2018-08-06 | 2019-01-04 | 中国地质大学（武汉） | A kind of automated construction method and system towards geophysics field knowledge mapping |
| CN109960810A (en) * | 2019-03-28 | 2019-07-02 | 科大讯飞(苏州)科技有限公司 | A kind of entity alignment schemes and device |
| CN110674274A (en) * | 2019-09-23 | 2020-01-10 | 中国农业大学 | Knowledge graph construction method for food safety regulation question-answering system |
| CN111475629A (en) * | 2020-03-31 | 2020-07-31 | 渤海大学 | Knowledge graph construction method and system for math tutoring question-answering system |
| CN111522961A (en) * | 2020-04-09 | 2020-08-11 | 武汉理工大学 | Attention mechanism and entity description based industrial map construction method |
| CN111552820A (en) * | 2020-04-30 | 2020-08-18 | 江河瑞通（北京）技术有限公司 | Water engineering scheduling data processing method and device |
| CN111639190A (en) * | 2020-04-30 | 2020-09-08 | 南京理工大学 | Medical knowledge map construction method |
| CN111737496A (en) * | 2020-06-29 | 2020-10-02 | 东北电力大学 | Power equipment fault knowledge map construction method |
Non-Patent Citations (4)
| Title |
|---|
| Large-scale Semantic Integration of Linked Data: A Survey;Michalis Mountantonakis 等;《ACM Computing Surveys (CSUR)》;1-40 * |
| Ontology matching and verification system using concept description translation;Ioan Alfred Letia 等;《Proceedings of the 2010 IEEE 6th International Conference on Intelligent Computer Communication and Processing》;51-58 * |
| 基于几何深度学习的知识图谱关键技术研究进展;杜博 等;《航空兵器》;1-10 * |
| 语义模型支持的战场环境数据集成方法;李鸿飞 等;《测绘科学技术学报》;632-636 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112559765A (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112559765B (en) | Semantic integration method for multi-source heterogeneous database | |
| CN109902145B (en) | Attention mechanism-based entity relationship joint extraction method and system | |
| Weber et al. | Textual case-based reasoning | |
| CN103544242B (en) | Microblog-oriented emotion entity searching system | |
| CN106951558B (en) | Data processing method of tax intelligent consultation platform based on deep search | |
| CN111597347B (en) | Knowledge embedding defect report reconstruction method and device | |
| CN113806563B (en) | Architect knowledge graph construction method for multi-source heterogeneous building humanistic historical material | |
| KR102259390B1 (en) | System and method for ensemble question-answering | |
| CN101661462A (en) | Four-layer structure Chinese text regularized system and realization thereof | |
| CN115599902B (en) | Oil-gas encyclopedia question-answering method and system based on knowledge graph | |
| Cuzzola et al. | Evolutionary fine-tuning of automated semantic annotation systems | |
| US10901959B2 (en) | System, method, and program for reconciling input datasets with a model ontology | |
| CN114564573A (en) | Academic cooperative relationship prediction method based on heterogeneous graph neural network | |
| CN117149974A (en) | Knowledge graph question-answering method for sub-graph retrieval optimization | |
| CN113515616B (en) | Task driving system based on natural language | |
| CN114356990A (en) | Base named entity recognition system and method based on transfer learning | |
| CN114218406A (en) | Transmission solution generation method and system based on transmission knowledge graph | |
| Mustafa et al. | Optimizing document classification: Unleashing the power of genetic algorithms | |
| CN111753151B (en) | Service recommendation method based on Internet user behavior | |
| Biswas et al. | A review on fundamentals of case-based reasoning and its recent application in different domains | |
| CN117474010A (en) | Power grid language model-oriented power transmission and transformation equipment defect corpus construction method | |
| CN116595125A (en) | Open domain question-answering method based on knowledge graph retrieval | |
| CN110309258A (en) | A kind of input checking method, server and computer readable storage medium | |
| CN114942981A (en) | Question-answer query method and device, electronic equipment and computer readable storage medium | |
| CN111582945A (en) | Financial market prediction method based on complex network |
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 |