CN114996423A - Water conservancy knowledge graph complex question-answering method based on subproblem pruning - Google Patents
Water conservancy knowledge graph complex question-answering method based on subproblem pruning Download PDFInfo
- Publication number
- CN114996423A CN114996423A CN202210605035.2A CN202210605035A CN114996423A CN 114996423 A CN114996423 A CN 114996423A CN 202210605035 A CN202210605035 A CN 202210605035A CN 114996423 A CN114996423 A CN 114996423A
- Authority
- CN
- China
- Prior art keywords
- complex
- question
- water conservancy
- entity
- answer
- 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
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/288—Entity relationship models
-
- 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
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06F16/33—Querying
- G06F16/338—Presentation of query results
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a water conservancy knowledge graph complex question-answering method based on sub-question pruning, which comprises the following steps of: embedding the water conservancy knowledge graph into a plurality of spaces, coding the complex problem and mapping the complex problem into an embedded space which is the same as the water conservancy knowledge graph to generate a candidate answer set; analyzing the syntactic structure and semantic structure of the complex problem and generating a dependency syntactic analysis tree, and decomposing the complex problem to obtain a problem description graph; identifying answer types of the subproblems in the question description graph, and pruning the candidate answer set by using the answer types; and calculating the score of each answer in the candidate answer set after pruning, and taking the answer entity with the highest score as a final answer to return to the user. The invention effectively relieves the problem of losing correct answers caused by the problem of sparse map, effectively reduces the range of candidate answer sets and obtains higher accuracy on complex questions and answers.
Description
Technical Field
The invention relates to computer software, in particular to a water conservancy knowledge graph complex question-answering method based on sub-problem pruning.
Background
According to the number of knowledge graph triples, natural language problems can be divided into two categories:
(1) simple questions, only depend on a triplet to finish the question and answer;
(2) complex problems, involving multiple classes and attributes of multiple ontologies or having multiple constraints, have complex semantic and syntactic structures.
Analyzing natural language questions input by a user based on complex questions and answers of a knowledge graph to obtain a question subject entity, linking the subject entity to the knowledge graph, finding a plurality of associated triples to model a multi-hop long path, and returning a tail entity of the path to the user as a question answer. At present, a complex question-answering system based on a knowledge graph is well applied to multiple vertical industries such as finance, medical treatment, e-commerce and the like, and accurate and standard industry data description and expression are provided for users by introducing the knowledge graph. However, the existing water conservancy knowledge graph has the problems that part of entities lack knowledge or relationship paths and the like, and the graph has sparsity. The question-answer derivation process depends on the connection of relations in the knowledge graph, and in the answer reasoning process, due to the sparsity of the graph, a path of a correct answer does not exist in a candidate subgraph formed in a limited hop count, so that the question posed by a user cannot be answered accurately.
Knowledge-graph-based complex question-answering generally requires consideration of a sub-graph of each question centered on the subject entity, but as the number of hops increases, the number of candidate answers grows exponentially. At present, the question answering of most complex questions adopts a sequential decision method. Qiu et al propose interpretable reasoning mechanism to gradually obtain answer entities of the questions, and on this basis, increase attention mechanism to ensure the accuracy of the reasoning process, and optimize the query path by using cluster search, thereby reducing the number of candidate answers. The method makes great progress in answering two-hop questions, but has poor performance in answering three-hop or complex-constraint questions, and causes error accumulation in the sequential decision process, thereby greatly limiting the ability of answering complex questions. Therefore, how to filter effectively is not related to the fact, and reducing the search space is the focus of research.
Disclosure of Invention
The invention aims to: the invention aims to provide a water conservancy knowledge graph complex question-answering method based on sub-question pruning, so that the problem that correct answers cannot be found due to incomplete knowledge graphs is relieved, and the question-answering accuracy is improved.
The technical scheme is as follows: the invention relates to a water conservancy knowledge graph complex question-answering method based on sub-question pruning, which comprises the following steps of:
(1) knowledge graph embedding and question embedding: embedding a knowledge map, namely embedding the water conservancy knowledge map into a plurality of spaces to capture comprehensive characteristic information; problem embedding, namely encoding a problem and mapping the problem to an embedding space which is the same as a water conservancy knowledge map; and calculating similarity scores of the question vectors and the entities in the knowledge graph to generate a candidate answer set.
(2) And (3) problem decomposition: the method comprises the steps of analyzing the syntactic structure and the semantic structure of the complex problem in detail, generating a corresponding dependency syntactic analysis tree, decomposing the complex problem according to the dependency syntactic analysis tree and a set of rules to obtain a problem description diagram, and displaying a series of simple sub-problems corresponding to the complex problem and the relationship of the sub-problems.
(3) Pruning of the candidate answer set: and identifying answer types of the sub-questions in the question description graph in the question decomposition step, and pruning all entities in the candidate answer set by using the answer types.
(4) And (3) answer screening: and calculating the score of each answer in the new candidate answer set, and taking the answer entity with the highest score as a final answer to be returned to the user.
The step (1) is specifically as follows:
(1.1) aiming at embedding of a knowledge graph, defining a head entity and a tail entity h, t epsilon E and a relation R epsilon R in the water conservancy knowledge graph, and learning vector representation of the head entity and the tail entity in a ComplEx space by utilizing a Complex method according to a scoring function
All reasonable triplets have phi (h, r, t)>0, all unreasonable triplets are Φ (h, r, t)<0。
(1.2) embedding the problems, coding the problems, embedding the problems into 768-dimensional vectors by using RoBERTA, mapping the complex problems to the complex space which is the same as the water conservancy knowledge map by 4 full-link layers and a ReLU activation function, and obtaining a sentence vector
(1.3) vector of question e by means of semantic scoring function in ComplEx q Replacing the relation r in the scoring function, and calculating e h ,e q ,e a To generate a candidate answer set a E.
The step (2) is specifically as follows:
and (2.1) carrying out detailed analysis on the syntactic structure and the semantic structure of the complex problem to generate a corresponding dependency syntactic analysis tree.
(2.2) identifying the problem type according to the dependency syntax analysis tree of the complex problem, and iteratively identifying the components of the sentence by combining a group of grammar rules to construct a problem description graph; each block in the problem description graph represents a sub-problem, where the ellipse-shaped nodes represent entities in the knowledge-graph, the rounded rectangular nodes describe the current entity node, and the dashed lines refer to edges connecting the description node to intermediate entities nested in the description.
The grammar rule in step (2.2) includes:
w-rule: the query words such as "which, where, which, what …" and a phrase describing the target entity;
c-rule: i.e. word connection rules;
a-rule: namely, the rules of definite language or idiom clauses;
n-rule: i.e., prepositional phrases or compound noun phrase rules.
The step (3) is specifically as follows:
and (3.1) firstly, constructing a relation-entity type mapping file according to the relation between the relation words and the entity types in the relation word definition of the water conservancy knowledge map.
(3.2) each block in the problem description diagram represents a subproblem, the rounded rectangle of each block describes the current entity, the similarity between the descriptions and twelve relation words contained in the water conservancy knowledge graph is calculated by using a score function S (d, r) in a PullNet model, and the relation words of S (d, r) >0.5 are reserved.
And (3.3) mapping the reserved relation according to the relation-entity type mapping rule to obtain the candidate answer type of the subproblem.
And (3.4) filtering all entities in the candidate answer set by using the candidate answer types of the subproblems to obtain a new candidate answer set A'.
In the step (4), the formula for calculating the score of each answer in the new candidate answer set is as follows:
wherein e is h As a vector representation of the head entity in the knowledge-graph, e q As a question vector, e a′ As entity vectors in the candidate answer set, e ans Is the final answer entity vector.
A computer storage medium having a computer program stored thereon, the computer program when executed by a processor implementing a water conservancy knowledge graph complex question-answering method based on sub-question pruning.
A computer device comprises a storage, a processor and a computer program which is stored on the storage and can run on the processor, wherein the processor executes the computer program to realize the water conservancy knowledge map complex question-answering method based on sub-question pruning.
Has the advantages that: compared with the prior art, the invention has the following advantages:
1. according to the method, the water conservancy knowledge map and the complex question are embedded into the same complex space, the semantic scoring function is adopted to calculate the correlation degree between the question and the entity in the map, the candidate answer set range is preliminarily screened out, and the problem of loss of correct answers caused by map sparse problems is effectively relieved;
2. the invention decomposes the complex question to obtain the question description graph, identifies the answer type of the sub-question in the question description graph, uses the answer type of the sub-question to prune the irrelevant fact in the candidate answer set, effectively reduces the range of the candidate answer set, reduces the search space, and obtains higher accuracy on the complex question and answer compared with the prior method.
Drawings
FIG. 1 is a flow chart of the steps of the present invention.
FIG. 2 is a problem embedding framework diagram.
FIG. 3 is a flow chart for generating a problem description graph.
FIG. 4 is an example diagram of a problem description graph.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
As shown in FIG. 1, a water conservancy knowledge graph complex question-answering method based on sub-question pruning comprises the following steps:
(1) embedding the water conservancy knowledge map and the complex questions proposed by the user into the same complex space, and calculating the semantic similarity of entity vectors and question vectors in the map to obtain an initial candidate answer set A.
(1.1) defining head entity and tail entity h, t epsilon E and relation R epsilon R in the water conservancy knowledge map, and learning vector representation of the head entity and the tail entity in a ComplEx space by utilizing a CompelEx method according to a scoring function
All reasonable triplets have phi (h, r, t)>0, all unreasonable triplets have phi (h, r, t)<0. The scoring function is as follows:
wherein Re represents the real part of an element,
denotes e t The conjugate vector of (2).
(1.2) for problem embedding, as shown in fig. 2, firstly, the problem is subjected to position coding and feature coding, the problem is embedded into a 768-dimensional vector by using a RoBERTA pre-training model to obtain a vector representation of the problem, and then the complex problem is mapped to a complex space which is the same as a water conservancy knowledge map by 4 full connection layers and a ReLU activation function to obtain a sentence vector
e q The method comprises the complex semantic relation of the question sentences.
(1.3) vector of question e with semantic scoring function in CompelEx q Replacing the relation r in the scoring function, and calculating e h ,e q ,e a To generate a candidate answer set a E. The calculation formula is as follows:
(2) and generating a dependency syntax analysis tree of the complex problem, and constructing a problem description diagram according to the dependency syntax analysis tree of the complex problem.
And (2.1) carrying out detailed analysis on the syntactic structure and the semantic structure of the complex problem to generate a corresponding dependency syntactic analysis tree.
(2.2) As shown in FIG. 3, according to the dependency parsing tree of the complex problem, first identify the problem type and determine the root node of the problem description graph. Initializing Block, regarding the whole problem as an entity description, detecting whether the entity description is matched with an A-rule or an N-rule, if so, indicating that a sub-fact exists in the entity description, and generating a new Block comprising a new entity node and a description node; if not, it is checked whether the entity description matches the W-rule or the C-rule, and if so, the entity description is decomposed into two entity descriptions pointing to the same entity. And detecting whether the newly generated entity description is matched with one of the four rules, and if so, executing the operation according to the operation until any rule is not met any more. For entity descriptions referencing other entities, a referencing edge is issued to point to the referenced entity, as in FIG. 4 entity description "# entry 1 flows through" a dashed referencing edge is issued to point to "entry 1". The specific grammar rules include:
w-rule: is composed of interrogative words such as "which, where, which, what …" and a phrase describing the target entity, e.g. "which rivers are more than 8000 km long? ";
c-rule: i.e. word connection rules;
a-rule: namely the rules of the idiom or the idiom clause;
n-rule: i.e., prepositional phrases or compound noun phrase rules, e.g., "river belonging to the water system of the Roc Bay".
(3) And identifying answer types of the sub-questions in the question description graph, and filtering all entities in the candidate answer set by using the answer types to generate a new candidate answer set A'.
And (3.1) firstly, constructing a relation-entity type mapping file according to the relation between the relation words and the entity types in the relation word definition of the water conservancy knowledge map.
(3.2) each block in the problem description graph represents a subproblem, a rounded rectangle of each block describes a subproblem answer entity, the similarity between the description D and twelve relation words contained in the water conservancy knowledge graph is calculated by using a score function S (D, r) in a PullNet model, and the relation words with the S (D, r) >0.5 are reserved. Wherein, the similarity is defined as that BERT _ FLAT of description represents a dot product embedded with the relation word r, the dot product is brought into the range of [0, 1] through a sigmoid function, and all relations are sorted. The calculation formula is as follows:
X d ＝BERT_FLAT(d)
and (3.3) mapping the reserved relation according to the relation-entity type mapping rule to obtain the candidate answer type of the subproblem.
And (3.4) pruning all entities in the candidate answer set by using the candidate answer types of the subproblems to obtain a new candidate answer set A'.
(4) Calculating the score of each answer in the candidate answer set A', taking the answer entity with the highest score as the final answer to return to the user, wherein the calculation formula is as follows:
wherein e is h As a vector representation of the head entity in the knowledge-graph, e q As a question vector, e a′ As entity vectors in the candidate answer set, e ans Is the final answer entity vector.
Claims (8)
1. A water conservancy knowledge graph complex question-answering method based on sub-question pruning is characterized by comprising the following steps:
(1) knowledge graph embedding and question embedding: embedding a knowledge map, namely embedding the water conservancy knowledge map into a plurality of spaces to capture comprehensive characteristic information; problem embedding, namely encoding a problem and mapping the problem to an embedding space which is the same as a water conservancy knowledge map; calculating similarity scores of the question vectors and entities in the knowledge graph to generate a candidate answer set;
(2) problem decomposition: analyzing the syntactic structure and semantic structure of the complex problem in detail, generating a corresponding dependency syntactic analysis tree, decomposing the complex problem according to the dependency syntactic analysis tree and a group of rules to obtain a problem description diagram, wherein the problem description diagram shows a series of simple sub-problems corresponding to the complex problem and the relationship of the sub-problems;
(3) pruning of the candidate answer set: identifying answer types of the subproblems in the question description graph in the question decomposition step, and pruning all entities in the candidate answer set by using the answer types;
(4) and (3) answer screening: and calculating the score of each answer in the new candidate answer set, and taking the answer entity with the highest score as a final answer to be returned to the user.
2. The water conservancy knowledge graph complex question-answering method based on subproblem pruning according to claim 1, wherein the step (1) is specifically as follows:
(1.1) aiming at embedding of a knowledge graph, defining a head entity and a tail entity h, t epsilon E and a relation R epsilon R in the water conservancy knowledge graph, and learning vector representation of the head entity and the tail entity in a ComplEx space by utilizing a Complex method according to a scoring function
All reasonable triplets have phi (h, r, t)>0, all unreasonable triplets are Φ (h, r, t)<0；
(1.2) embedding the problems, coding the problems, embedding the problems into 768-dimensional vectors by using RoBERTA, mapping the complex problems to complex spaces the same as the water conservancy knowledge maps by 4 full-connection layers and ReLU activation functions to obtain a sentence vector
(1.3) vector of question e by means of semantic scoring function in ComplEx q Replacing the relation r in the scoring function, and calculating e h ,e q ,e a To generate a candidate answer set a E.
3. The water conservancy knowledge graph complex question-answering method based on subproblem pruning according to claim 1, wherein the step (2) is specifically as follows:
(2.1) carrying out detailed analysis on the syntactic structure and the semantic structure of the complex problem to generate a corresponding dependency syntactic analysis tree;
(2.2) identifying the problem type according to the dependency syntax analysis tree of the complex problem, and iteratively identifying the components of the sentence by combining a group of grammar rules to construct a problem description graph; each block in the problem description graph represents a sub-problem, where the ellipse-shaped nodes represent entities in the knowledge-graph, the rounded rectangular nodes describe the current entity node, and the dashed lines refer to edges connecting the description node to intermediate entities nested in the description.
4. The water conservancy knowledge graph complex question-answering method based on sub-question pruning according to claim 3, wherein the grammatical rules in step (2.2) comprise:
w-rule: composed of interrogative words such as "which, where, which, what …" and a phrase describing the target entity;
c-rule: i.e. word connection rules;
a-rule: namely the rules of the idiom or the idiom clause;
n-rule: i.e., prepositional phrase or compound noun phrase rules.
5. The water conservancy knowledge graph complex question-answering method based on subproblem pruning according to claim 1, wherein the step (3) is specifically as follows:
(3.1) firstly, establishing a relation-entity type mapping file according to the relation between the relation words and the entity types in the relation word definition of the water conservancy knowledge map;
(3.2) each block in the problem description diagram represents a subproblem, the rounded rectangles of each block describe the current entity, the similarity between the descriptions and twelve relation words contained in the water conservancy knowledge graph is calculated by using a score function S (d, r) in a PullNet model, and the relation words of S (d, r) >0.5 are reserved;
(3.3) mapping the reserved relation according to the relation-entity type mapping rule to obtain a candidate answer type of the subproblem;
and (3.4) filtering all entities in the candidate answer set by using the candidate answer types of the subproblems to obtain a new candidate answer set A'.
6. The water conservancy knowledge graph complex question-answering method based on sub-question pruning according to claim 1, wherein in the step (4), the formula for calculating the score of each answer in the new candidate answer set is as follows:
wherein e is h As a vector representation of the head entity in the knowledge-graph, e q As a question vector, e a ' As entity vector in candidate answer set, e ans Is the final answer entity vector.
7. A computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a sub-problem pruning-based hydraulic knowledgebase complex question-answering method according to any one of claims 1 to 6.
8. A computer device comprising a storage, a processor and a computer program stored on the storage and executable on the processor, wherein the processor implements the sub-problem pruning based water conservancy knowledge graph complex question-answering method according to any one of claims 1 to 6 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210605035.2A CN114996423A (en) | 2022-05-31 | 2022-05-31 | Water conservancy knowledge graph complex question-answering method based on subproblem pruning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210605035.2A CN114996423A (en) | 2022-05-31 | 2022-05-31 | Water conservancy knowledge graph complex question-answering method based on subproblem pruning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114996423A true CN114996423A (en) | 2022-09-02 |
Family
ID=83031928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210605035.2A Pending CN114996423A (en) | 2022-05-31 | 2022-05-31 | Water conservancy knowledge graph complex question-answering method based on subproblem pruning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114996423A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116089587A (en) * | 2023-02-20 | 2023-05-09 | 星环信息科技(上海)股份有限公司 | Answer generation method, device, equipment and storage medium |
| CN116821312A (en) * | 2023-08-30 | 2023-09-29 | 中国石油大学(华东) | Complex question-answering method based on discipline field knowledge graph |
| CN117350387A (en) * | 2023-12-05 | 2024-01-05 | 中水三立数据技术股份有限公司 | Intelligent question-answering system based on water conservancy knowledge platform |
-
2022
- 2022-05-31 CN CN202210605035.2A patent/CN114996423A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116089587A (en) * | 2023-02-20 | 2023-05-09 | 星环信息科技(上海)股份有限公司 | Answer generation method, device, equipment and storage medium |
| CN116089587B (en) * | 2023-02-20 | 2024-03-01 | 星环信息科技(上海)股份有限公司 | Answer generation method, device, equipment and storage medium |
| CN116821312A (en) * | 2023-08-30 | 2023-09-29 | 中国石油大学(华东) | Complex question-answering method based on discipline field knowledge graph |
| CN116821312B (en) * | 2023-08-30 | 2023-11-14 | 中国石油大学(华东) | Complex question-answering method based on discipline field knowledge graph |
| CN117350387A (en) * | 2023-12-05 | 2024-01-05 | 中水三立数据技术股份有限公司 | Intelligent question-answering system based on water conservancy knowledge platform |
| CN117350387B (en) * | 2023-12-05 | 2024-04-02 | 中水三立数据技术股份有限公司 | Intelligent question-answering system based on water conservancy knowledge platform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA3055379C (en) | Automated tool for question generation | |
| Seo et al. | Solving geometry problems: Combining text and diagram interpretation | |
| CN106844368B (en) | Method for man-machine conversation, neural network system and user equipment | |
| CN114996423A (en) | Water conservancy knowledge graph complex question-answering method based on subproblem pruning | |
| CN110245238B (en) | Graph embedding method and system based on rule reasoning and syntax mode | |
| CN111241840A (en) | Named entity identification method based on knowledge graph | |
| CN112328800A (en) | System and method for automatically generating programming specification question answers | |
| CN115099219A (en) | Aspect level emotion analysis method based on enhancement graph convolutional neural network | |
| CN115757715A (en) | Complex problem multi-hop intelligent question-answering method based on knowledge graph representation learning | |
| CN113705196A (en) | Chinese open information extraction method and device based on graph neural network | |
| Cao et al. | Relmkg: reasoning with pre-trained language models and knowledge graphs for complex question answering | |
| Liu et al. | Vera: A general-purpose plausibility estimation model for commonsense statements | |
| CN112417170B (en) | Relationship linking method for incomplete knowledge graph | |
| Lien et al. | Semantic parsing for textual entailment | |
| Cheng et al. | Building a neural semantic parser from a domain ontology | |
| CN117648984A (en) | Intelligent question-answering method and system based on domain knowledge graph | |
| CN111651528A (en) | Open entity relation extraction method based on generative countermeasure network | |
| CN113139657B (en) | Machine thinking realization method and device | |
| Datta et al. | Optimization of an automated examination generation system using hybrid recurrent neural network | |
| Pilato et al. | Sub-symbolic semantic layer in cyc for intuitive chat-bots | |
| CN115392257A (en) | Text relation extraction method based on graph aggregation and reasoning | |
| CN114443818A (en) | Dialogue type knowledge base question-answer implementation method | |
| CN114398905A (en) | Crowd-sourcing-oriented problem and solution automatic extraction method, corresponding storage medium and electronic device | |
| Crouse | Question-Answering with Structural Analogy | |
| Wang | An improved knowledge graph question answering system for English teaching |
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 |