CN115563297A - Food safety knowledge graph construction and completion method based on graph neural network - Google Patents
Food safety knowledge graph construction and completion method based on graph neural network Download PDFInfo
- Publication number
- CN115563297A CN115563297A CN202211134812.6A CN202211134812A CN115563297A CN 115563297 A CN115563297 A CN 115563297A CN 202211134812 A CN202211134812 A CN 202211134812A CN 115563297 A CN115563297 A CN 115563297A
- Authority
- CN
- China
- Prior art keywords
- graph
- entity
- food safety
- representation
- knowledge graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Machine Translation (AREA)
Abstract
The invention provides a food safety knowledge graph building and complementing method based on a graph neural network, which comprises the steps of acquiring national food safety standard files related to food types, food additives and pesticide residues in food, processing the national food safety standard files into triples capable of being applied to the knowledge graph through operations such as data cleaning, formatting and the like, and building a body layer mode framework of the food safety knowledge graph; designing a food safety knowledge map query system capable of realizing entity query and visual display; generating a word vector of an entity name in the food safety knowledge graph according to a pre-training language model BERT; and performing feature fusion on text information and graph structure information in the food safety knowledge graph by using a graph neural network architecture, and respectively applying the feature fusion to two downstream tasks of entity classification and link prediction so as to fulfill the aim of knowledge graph completion. The invention improves the integrity and the practicability of the food safety knowledge map and realizes the intelligent application of food safety standard information.
Description
Technical Field
The invention belongs to the technical field of knowledge maps and graph neural networks, and particularly relates to a food safety knowledge map construction and completion method based on a graph neural network.
Background
Along with the continuous improvement of the living standard of people, the attention of people to food safety is higher and higher. Whether in the supervision department or the general public, a set of practical and effective standard guidance is needed. At present, however, the national food safety standard is still stored in the form of files, and the files are various in types and number, not uniform in format, and have a reference relationship with each other. Although the work of food quality standardization and informatization management in China is continuously promoted, the method is still lack of structural integration of the standards and systematic analysis association.
With the development of scientific technology, the knowledge graph can describe the concepts and the relations of the objective world as nodes and edges in a graph structure form, and is more beneficial to inquiring the contents of food standards and the relations among the food standards. At present, most food safety knowledge maps at home and abroad are concentrated on the construction of sub-fields, and an authoritative knowledge base of food safety standards is not used as guidance.
If the knowledge graph is constructed by only manually extracting standard contents as data, the knowledge graph is definitely inefficient for the huge field of food, and the information is often incomplete. The graph neural network has the advantage of processing graph structure data and can be applied to entity classification and link prediction tasks of knowledge graph completion in an attempt mode. However, due to the structure of the graph neural network itself, an over-smoothing phenomenon occurs when deep graph information is acquired, which affects the effect of the model, and if only the graph structure information is considered, the information of the knowledge graph itself is not fully utilized, which is one of the problems to be considered today.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the food safety knowledge graph construction and completion method based on the graph neural network is used for improving the integrity of the food safety knowledge graph.
The technical scheme adopted by the invention for solving the technical problems is as follows: a food safety knowledge graph construction and completion method based on a graph neural network comprises the following steps:
(in accordance with the claims herein, filled after validation of the claims)
The beneficial effects of the invention are as follows:
1. the invention relates to a food safety knowledge graph construction and completion method based on a graph neural network, which comprises the steps of constructing a knowledge graph by extracting food safety standard information and generating word vectors of entity names in the food safety knowledge graph according to a pre-training language model BERT; text information and graph structure information in the food safety knowledge graph are subjected to feature fusion by utilizing a graph neural network architecture, and are respectively applied to two downstream tasks of entity classification and link prediction to complement the food safety knowledge graph, so that the function of improving the integrity of the food safety knowledge graph is realized; according to the method, the text representation is introduced into the graph neural network, so that node information is enriched, an over-smooth phenomenon caused by the self structure of the multilayer graph neural network is relieved, and a model can learn more knowledge map information; the invention improves the scoring function in the R-GCN link prediction model, introduces the scoring function constructed by the idea of sectional block dot product, and improves the link prediction effect in the form of respective calculation of the graph structure characteristic vector and the text characteristic vector.
2. According to the method, the national food safety standard files related to food categories, food additives and pesticide residues in food are obtained, the food safety national standard information such as the maximum quantity of the pesticide residues in the food, different food categories and food additives is processed into a triple form capable of being applied to the knowledge graph through operations such as data cleaning and formatting, the body layer mode framework of the food safety knowledge graph is constructed, and the vacancy of knowledge graph data in the food safety field is made up.
3. The invention builds a Python-based flash frame and Echarts.js webpage end visualization system, stores the data of the national standard content of food safety in the form of a graph and applies the data to downstream tasks, thereby realizing the food safety knowledge map query system with entity query and visualization display.
4. The invention makes up the defects of the prior art, improves the integrity and the practicability of the food safety knowledge graph, and realizes the intelligent application of food safety standard information.
Drawings
FIG. 1 is a flow chart of a preliminary construction of a food safety knowledgebase map in an embodiment of the invention.
Fig. 2 is an ontology structure diagram of the food safety knowledge-graph according to the embodiment of the present invention.
FIG. 3 is a flow diagram of data extraction and processing according to an embodiment of the invention.
FIG. 4 is a schematic structural diagram of an entity classification model according to an embodiment of the present invention.
Fig. 5 is a schematic structural diagram of a link prediction model according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, an embodiment of the invention comprises a preliminary construction module of a food safety knowledge graph, a knowledge graph entity classification model based on a graph neural network and a text pre-training model, and a knowledge graph link prediction model based on a coder-decoder.
The invention relates to a food safety knowledge graph construction and completion method based on a graph neural network, which comprises the following steps of:
s1: collecting data; extracting national standard information of food safety by means of web crawler and table analysis, wherein the national standard information comprises semi-structured information and unstructured information in a standard document; the work flow diagram of data extraction is shown in fig. 3;
s11: acquiring information of the national food safety standard from a website by a web crawler method, and downloading a national food safety standard file;
s12: OCR identifies the downloaded national standard file of food safety, and analyzes the table information in the file;
s13: and obtaining semi-structured data and unstructured data based on rule matching and webpage and table analysis.
S2: constructing an ontology; based on the obtained semi-structured information and manual auxiliary labeling, an ontology framework of the food safety knowledge graph is constructed, the obtained ontology framework is utilized to further improve the data extraction mode, and the obtained ontology framework graph is shown as a graph in fig. 2, wherein the graph comprises entity types and relationship types related to food products, food standards, food additives and pesticide residues;
s3: processing data; formatting and cleaning the semi-structured and unstructured data according to the storage requirements of the constructed body frame and the knowledge graph to obtain triple data in a csv format; the work flow diagram of data processing is shown in FIG. 3;
s4: storing the knowledge graph and constructing a visual query system;
s41: storing the three-element data of the knowledge graph in the csv format into a Neo4J graph database;
s42: calling a Neo4j graph database based on a flash framework of Python, and displaying a knowledge graph by a force directed graph of Echarts. Js;
s43: and calling a cypher statement of Neo4J through Python to realize the query of the knowledge graph entity, and displaying the knowledge graph entity on a webpage.
S5: completing a knowledge graph entity classification task by using a graph neural network and a text pre-training model BERT; FIG. 4 shows a schematic diagram of a convolution network of a relational graph combined with text representation in the invention, which comprises a BERT text representation part and an R-GCN graph structure information part;
s51: constructing a data set based on the triples of the food safety knowledge graph, dividing a training set, a verification set and a test set, setting the entity types related to food safety as corresponding type numbers, abstracting the entities as entity numbers, and constructing a DGLDataset data set based on an entity type table and a triad table;
s52: adopting a BERT pre-training model as an entity classification model (see figure 4), extracting text word vector representation of the entity name of the knowledge graph, and performing dimension reduction processing on word vectors obtained by pre-training;
the entity classification model is used for expanding the knowledge graph representation from single network structure representation to common representation of structure and text by combining the structure information of the graph and the text description of the entity, and constructing a model fusing multi-source knowledge graph information; obtaining word vector representation of entity description through a text pre-training model BERT, adding the word vector representation into the entity representation, fusing the word vector representation with original entity relation structure information, and learning knowledge graph representation through a relation graph convolution network model;
the BERT service is adopted, and the pretrained BERT model is packaged based on the C/S architecture and is provided to the client as a service, so that support is provided for engineering projects more conveniently. As most of the entity names in the food safety knowledge map are Chinese vocabularies, a Chinese BERT pre-training model, namely Chinese _ L-12_H-768 _A-12, is selected to obtain 768-dimensional word vectors of all the entity names. Because the word vector obtained by pre-training the model has higher dimensionality, if the word vector is directly used, the calculated amount of the model is greatly increased, and the 768-dimensional word vector with higher dimensionality is subjected to dimensionality reduction by a Principal Component Analysis (PCA) method. Subsequent experiments prove that the reduction of the dimension has little influence on experimental results, the dimension of the text representation uniformly used by the entity classification model is 5 dimensions and 4 decimal places are reserved for floating point numbers in consideration of the computational complexity and the completeness of text information representation, and the subsequent use is facilitated.
S53: updating the node representation by utilizing the R-GCN aggregation information to acquire graph structure information;
s54: embedding a text word vector of an entity name in front of an output layer of the R-GCN;
s55: after learning of an output layer, entities are classified by using softmax, possible classes are predicted for each entity in a sample, and a minimum cross entropy loss function is optimized on all labeled nodes.
Experiments prove that the R-GCN model combined with the text information obtains higher entity classification accuracy when comparing whether text representation is used or not.
S6: completing a knowledge graph link prediction task based on a graph neural network model of an encoder-decoder; FIG. 5 illustrates a block diagram of an encoder-decoder based link prediction model, including a BERT textual representation portion and an R-GCN graph structure information portion; taking the R-GCN model as an encoder to obtain the representation of the triple, and taking the scoring function as a decoder;
the entity classification is realized through the knowledge graph constructed by the triples and based on the graph neural network and the text representation, the integrity of the food safety knowledge graph on the entity type is improved, but the food safety knowledge graph has the problem of partial relation loss. Whether the triple meets the requirement or not is judged based on the scoring function by predicting the missing triple in the knowledge graph, so that the link prediction is realized. For the triples in the form of (head entity, relation and tail entity), predicting another entity according to one entity and the corresponding relation, thereby solving the problem of head and tail entity prediction.
The link prediction model obtained in this way can also be applied to inference of knowledge graphs, predicting other entities that may have some specific association with it based on one entity and its relationship structure. Regarding the food safety problem, other foods having a certain relation with the food or other substances which may exceed the standard can be deduced according to a certain food category through the method, so that potential safety hazards which may exist in the whole chain (production, transportation, storage and the like) of the food can be accurately grasped, the problem can be timely found, and the problem can be prevented.
In the link prediction task, two aspects of work are mainly involved. On the one hand, in order for the model to learn the knowledge of the knowledge-graph, a representation of the knowledge-graph needs to be obtained. The type and representation of edges in the graph need to be considered more than in the entity classification model. On the other hand, a triple scoring function suitable for the current problem needs to be selected, and the function directly influences the training and the result of the model.
S61: dividing a food safety knowledge map data set, avoiding the influence of type imbalance on the prediction effect, ensuring that the relation type proportion of a training set, a verification set and a test set is basically consistent as much as possible when data is extracted, and abstracting the entity type and the relation type related to food safety into corresponding type numbers;
s62: performing text feature extraction on the description information of each node in the knowledge graph by using a BERT pre-training model to obtain word vector representation;
s63: the encoder portion of the linked prediction model: updating the representation of the nodes based on the hidden layer of the R-GCN, wherein the description of the nodes is used as text information to obtain word vector representation through a BERT pre-training model, embedding the text representation word vector in front of the last hidden layer of the R-GCN, and inputting the graph structure information of the knowledge graph to the hidden layer of the R-GCN after initialization embedding to obtain the representation of the knowledge graph; the two characteristics are fused and represented and learned before entering the last hidden layer, and then pass through the last R-GCN hidden layer;
s64: decoder part of the link prediction model: and evaluating the triples through a scoring function to obtain the ranking of the positive samples in all sample scores. The scoring function is defined as f (s, r, o), where s, r, o represent the head, relationship and tail entities, respectively.
S641: the scoring function is divided into f by combining graph structure information and text information contained in node representation based on the segmented block dot product idea graph (s, r, o) and f word (s, r, o), i.e., f (s, r, o) = f graph (s，r，o)+f word (s, r, o), dividing the graph structure feature and the text feature into 2 parts, f graph (s，r，o)＝∑ 0≤x，y＜2 s x，y ·<r x ，s y ，o z >，f word (s，r，o)＝∑ 2≤x，y＜4 <r x ，s y ，o z >(ii) a Where x, y, z represent segment numbers of the relational representation, the head entity representation, and the tail entity representation, respectively.
S642: the graph structure information part needs to consider symmetry, an odd number is used for representing asymmetric relation, an even number is used for representing symmetric relation, and a parameter s is introduced x，y And no relation constraint is added in the calculation of the word vector part. When x is odd number and x + y is greater than or equal to 2, s x，y Is-1 and the rest is 1.
S643: to simplify the calculation, an index w of the tail entity is introduced x，y And the complexity is reduced.
S65: the data set constructed in the early stage does not contain negative example data, so that negative sampling is carried out by adopting a method of randomly destroying positive example data, namely, negative samples with a certain proportion are generated in a manner that other random entities in the data set replace head entities or tail entities, and optimization is carried out through cross entropy loss.
The above embodiments are only used for illustrating the design idea and features of the present invention, and the purpose of the present invention is to enable those skilled in the art to understand the content of the present invention and implement it accordingly, and the protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes and modifications made in accordance with the principles and concepts disclosed herein are intended to be included within the scope of the present invention.
Claims (9)
1. A food safety knowledge graph construction and completion method based on a graph neural network is characterized by comprising the following steps: the method comprises the following steps:
s1: extracting food safety national standard information by means of web crawler and table analysis, wherein the food safety national standard information comprises semi-structured data and unstructured data in a standard file;
s2: based on the semi-structured data, carrying out manual auxiliary labeling, constructing an ontology framework of the food safety knowledge graph, and further improving the data extraction mode by using the ontology framework;
s3: formatting and cleaning the semi-structured data and the unstructured data according to the storage requirements of the body frame and the knowledge graph to obtain three groups of data of the food safety knowledge graph in a csv format;
s4: storing three sets of data of food safety knowledge maps and constructing a visual query system;
s5: the convolution network of the relational graph is represented by combining text, and comprises a BERT text representation part and an R-GCN graph structure information part; adopting a graph neural network and a text pre-training model BERT to build an entity classification model and completing a knowledge graph entity classification task; the entity classification model is used for expanding the knowledge graph representation from single network structure representation to common representation of structure and text by combining the structure information of the graph and the text description of the entity, and constructing a model fusing multi-source knowledge graph information; obtaining word vector representation of entity description through a text pre-training model BERT, adding the word vector representation into the entity representation, fusing the word vector representation with original entity relationship structure information, and learning knowledge graph representation through a relationship graph convolution network model;
s6: taking the R-GCN model as an encoder to obtain the representation of the triple, and taking the scoring function as a decoder; constructing a structure of a coder-decoder based link prediction model, including a BERT text representation part and an R-GCN graph structure information part; the encoder-decoder based link prediction model is used for predicting other entities which may have a certain specific association with one entity according to the entity and the relation structure of the entity, and judging whether the triples meet the requirements or not based on the scoring function by predicting the missing triples in the knowledge graph, thereby completing the knowledge graph link prediction task.
2. The method for constructing and complementing the food safety knowledge graph based on the graph neural network as claimed in claim 1, wherein the method comprises the following steps: in the step S1, the specific steps are as follows:
s11: acquiring information of national food safety standards from a network through a crawler program, and downloading national food safety standard files;
s12: OCR identifies the downloaded national standard file of food safety, and analyzes the table information in the file;
s13: and obtaining semi-structured data and unstructured data based on rule matching and webpage and table analysis.
3. The method for constructing and complementing the food safety knowledge graph based on the graph neural network as claimed in claim 1, wherein the method comprises the following steps: in the step S2, the body frame includes entity types and relationship types related to food types, food standards, food additives, and pesticide residues.
4. The food safety knowledge domain building and complementing method based on the graph neural network as claimed in claim 3, wherein: in the step S4, the specific steps are as follows:
s41: storing the three-element data of the knowledge graph in the csv format into a Neo4J graph database;
s42: calling a Neo4j graph database based on a flash framework of Python, and displaying a knowledge graph by a force directed graph of Echarts. Js;
s43: and calling a cypher statement of Neo4J through Python to realize the query of the knowledge graph entity, and displaying the knowledge graph entity on a webpage.
5. The food safety knowledge domain building and complementing method based on the graph neural network as claimed in claim 4, wherein: in the step S5, the specific steps are as follows:
s51: constructing a data set based on the ternary group data of the food safety knowledge map, dividing a training set, a verification set and a test set, taking the entity type related to food safety as a corresponding type number, abstracting the entity to be the entity number, and constructing a DGLDataset data set based on the entity type table and the ternary group table;
s52: adopting a BERT pre-training model as an entity classification model, extracting text word vector representation of the entity name of the knowledge graph, and performing dimension reduction processing on word vectors obtained by pre-training;
s53: updating the node representation by utilizing the R-GCN aggregation information to acquire graph structure information;
s54: embedding a text word vector of an entity name in front of an output layer of the R-GCN;
s55: after output layer learning, entities are classified using softmax, possible classes are predicted for each entity in the sample, and optimization is performed by minimizing the cross entropy loss function over all labeled nodes.
6. The food safety knowledge domain building and complementing method based on the graph neural network as claimed in claim 5, wherein: in step S6, the entity classification model includes an input layer, a plurality of hidden layers, and an output layer;
each layer of the entity classification model comprises a relation graph convolution network layer R-GCN, and the relation graph convolution network layer R-GCN is used for calculating output information of each node on the training set according to the node representation and the type of the edge through a message function and aggregating the information to obtain new node representation;
the input layer of the model uses the type and number of the entity as features; the input layer and the hidden layer adopt ReLU as an activation function; the output of the last layer uses softmax for classification;
in the hidden layer, the node representation obtained by the input layer is calculated by a plurality of relational graph convolution network layers R-GCN and an activation function ReLU respectively; let Y be the index set of the node, K be the total number of categories,
predicted value of kth class, t, representing ith node labeled at l-th layer ik Representing a real label, the minimum cross entropy Loss function Loss obtained by the output layer on all labeled nodes is:
in the entity classification task, the calculated amount of the model is reduced in a basis function decomposition mode; and embedding the word vector obtained by the BERT pre-training model into an output layer of the R-GCN model for obtaining the best effect.
7. The food safety knowledge domain building and complementing method based on the graph neural network as claimed in claim 5, wherein: in the step S6, the specific steps are as follows:
s61: dividing a food safety knowledge map data set, and enabling the relation type proportion of a training set, a verification set and a test set to be consistent when data are extracted so as to avoid the influence of type imbalance on the prediction effect; abstracting the entity type and the relation type related to food safety into corresponding type numbers;
s62: performing text feature extraction on the description information of each node in the knowledge graph by using a BERT pre-training model, and obtaining word vector representation;
s63: an encoder part of the link prediction model updates the representation of nodes based on hidden layers of the R-GCN, obtains word vector representation by taking the description including the nodes as text information through a BERT pre-training model, embeds the text representation word vectors in front of the last hidden layer of the R-GCN, and obtains the representation of the knowledge graph by inputting the graph structure information of the knowledge graph into the hidden layers of the R-GCN after initializing and embedding; the two characteristics are fused and represented and learned before entering the last hidden layer, and then pass through the last R-GCN hidden layer;
s64: let s, r, o respectively represent head entity, relation and tail entity, and define a scoring function as f (s, r, o); the decoder part of the link prediction model realizes the evaluation of the triples through a scoring function to obtain the ranking of the positive samples in all sample scores;
s65: the data set constructed in the early stage does not contain negative example data, negative sampling is carried out by adopting a method of randomly destroying positive example data, negative samples with a certain proportion are generated in a mode that other random entities in the data set replace head entities or tail entities, and optimization is carried out through cross entropy loss.
8. The method for constructing and complementing the food safety knowledge graph based on the graph neural network as claimed in claim 7, wherein: in the step S64, the specific steps are:
s641: setting the graph structure characteristic of the scoring function as f graph (s, r, o) text feature f word (s, r, o), then:
f(s，r，o)＝f graph (s，r，o)+f word (s，r，o)；
let x, y, z be the segment number of relation representation, head entity representation and tail entity representation segmentation respectively, divide picture structure characteristic and text characteristic into 2 parts respectively, do:
f graph (s，r，o)＝∑ 0≤x，y＜2 s x，y ·<r x ，s y ，o z >，
f word (s，r，o)＝∑ 2≤x，y＜4 <r x ，s y ，o z >；
s642: the graph structure information part considers symmetry, uses odd numbers to represent asymmetric relation, even numbers to represent symmetric relation, and introduces a parameter s x，y (ii) a No relation constraint is added in the calculation of the word vector part; when x is odd number and x + y is greater than or equal to 2, s x，y Is-1, and the rest is 1;
s643: index w of imported tail entities x，y The method is used for simplifying calculation and reducing complexity.
9. A computer storage medium, characterized in that: stored therein is a computer program executable by a computer processor, the computer program performing a graph neural network based food safety knowledgegraph construction and completion method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211134812.6A CN115563297A (en) | 2022-09-19 | 2022-09-19 | Food safety knowledge graph construction and completion method based on graph neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211134812.6A CN115563297A (en) | 2022-09-19 | 2022-09-19 | Food safety knowledge graph construction and completion method based on graph neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115563297A true CN115563297A (en) | 2023-01-03 |
Family
ID=84740135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211134812.6A Pending CN115563297A (en) | 2022-09-19 | 2022-09-19 | Food safety knowledge graph construction and completion method based on graph neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115563297A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116186295A (en) * | 2023-04-28 | 2023-05-30 | 湖南工商大学 | Attention-based knowledge graph link prediction method, attention-based knowledge graph link prediction device, attention-based knowledge graph link prediction equipment and attention-based knowledge graph link prediction medium |
| CN116955648A (en) * | 2023-07-19 | 2023-10-27 | 上海企卓元科技合伙企业(有限合伙) | Knowledge graph analysis method based on non-privacy data association |
| CN116992959A (en) * | 2023-06-02 | 2023-11-03 | 广州数说故事信息科技有限公司 | Knowledge graph-based food product creative concept generation and verification method |
| CN117669718A (en) * | 2023-12-05 | 2024-03-08 | 广州鸿蒙信息科技有限公司 | Fire control knowledge training model and training method based on artificial intelligence |
| CN117932089A (en) * | 2024-03-25 | 2024-04-26 | 南京中医药大学 | Knowledge graph-based data analysis method |
-
2022
- 2022-09-19 CN CN202211134812.6A patent/CN115563297A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116186295A (en) * | 2023-04-28 | 2023-05-30 | 湖南工商大学 | Attention-based knowledge graph link prediction method, attention-based knowledge graph link prediction device, attention-based knowledge graph link prediction equipment and attention-based knowledge graph link prediction medium |
| CN116992959A (en) * | 2023-06-02 | 2023-11-03 | 广州数说故事信息科技有限公司 | Knowledge graph-based food product creative concept generation and verification method |
| CN116992959B (en) * | 2023-06-02 | 2024-03-15 | 广州数说故事信息科技有限公司 | Knowledge graph-based food product creative concept generation and verification method |
| CN116955648A (en) * | 2023-07-19 | 2023-10-27 | 上海企卓元科技合伙企业(有限合伙) | Knowledge graph analysis method based on non-privacy data association |
| CN116955648B (en) * | 2023-07-19 | 2024-01-26 | 上海企卓元科技合伙企业(有限合伙) | Knowledge graph analysis method based on non-privacy data association |
| CN117669718A (en) * | 2023-12-05 | 2024-03-08 | 广州鸿蒙信息科技有限公司 | Fire control knowledge training model and training method based on artificial intelligence |
| CN117932089A (en) * | 2024-03-25 | 2024-04-26 | 南京中医药大学 | Knowledge graph-based data analysis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2023065545A1 (en) | Risk prediction method and apparatus, and device and storage medium | |
| CN115563297A (en) | Food safety knowledge graph construction and completion method based on graph neural network | |
| CN112100398B (en) | Patent blank prediction method and system | |
| CN112183056A (en) | Context-dependent multi-classification emotion analysis method and system based on CNN-BilSTM framework | |
| Sridevi et al. | AI based suitability measurement and prediction between job description and job seeker profiles | |
| CN113254675B (en) | Knowledge graph construction method based on self-adaptive few-sample relation extraction | |
| CN112926308B (en) | Method, device, equipment, storage medium and program product for matching text | |
| CN114330966A (en) | Risk prediction method, device, equipment and readable storage medium | |
| CN117033571A (en) | Knowledge question-answering system construction method and system | |
| Rohit et al. | Machine intelligence based personality prediction using social profile data | |
| CN115545558A (en) | Method, device, machine readable medium and equipment for obtaining risk identification model | |
| CN115221387A (en) | Enterprise information integration method based on deep neural network | |
| Zhang et al. | Modeling and prediction of stock price with convolutional neural network based on blockchain interactive information | |
| CN111709225A (en) | Event cause and effect relationship judging method and device and computer readable storage medium | |
| Addepalli et al. | A proposed framework for measuring customer satisfaction and product recommendation for ecommerce | |
| Mishra et al. | Twitter sentiment analysis using naive bayes algorithm | |
| Ayala et al. | A neural network for semantic labelling of structured information | |
| CN117390198A (en) | Method, device, equipment and medium for constructing scientific and technological knowledge graph in electric power field | |
| CN111951079A (en) | Credit rating method and device based on knowledge graph and electronic equipment | |
| CN111815426A (en) | Data processing method and terminal related to financial investment and research | |
| CN111581379A (en) | Automatic composition scoring calculation method based on composition question-deducting degree | |
| CN116595191A (en) | Construction method and device of interactive low-code knowledge graph | |
| US20170228439A1 (en) | Automatic natural language processing based data extraction | |
| CN113239143B (en) | Power transmission and transformation equipment fault processing method and system fusing power grid fault case base | |
| CN113869034B (en) | Aspect emotion classification method based on reinforced dependency graph |
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 |