CN113051382A - Intelligent power failure question-answering method and device based on knowledge graph - Google Patents
Intelligent power failure question-answering method and device based on knowledge graph Download PDFInfo
- Publication number
- CN113051382A CN113051382A CN202110377904.6A CN202110377904A CN113051382A CN 113051382 A CN113051382 A CN 113051382A CN 202110377904 A CN202110377904 A CN 202110377904A CN 113051382 A CN113051382 A CN 113051382A
- Authority
- CN
- China
- Prior art keywords
- module
- question
- semantic
- knowledge
- power failure
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000013507 mapping Methods 0.000 claims abstract description 33
- 238000000605 extraction Methods 0.000 claims description 25
- 238000010606 normalization Methods 0.000 claims description 15
- 238000004458 analytical method Methods 0.000 claims description 9
- 238000013136 deep learning model Methods 0.000 claims description 9
- 230000003993 interaction Effects 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000003190 augmentative effect Effects 0.000 claims 1
- 239000004557 technical material Substances 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 206010063385 Intellectualisation Diseases 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- COCAUCFPFHUGAA-MGNBDDOMSA-N n-[3-[(1s,7s)-5-amino-4-thia-6-azabicyclo[5.1.0]oct-5-en-7-yl]-4-fluorophenyl]-5-chloropyridine-2-carboxamide Chemical compound C=1C=C(F)C([C@@]23N=C(SCC[C@@H]2C3)N)=CC=1NC(=O)C1=CC=C(Cl)C=N1 COCAUCFPFHUGAA-MGNBDDOMSA-N 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—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
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3338—Query expansion
-
- 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
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Human Computer Interaction (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application provides a power failure intelligent question-answering method and device based on a knowledge graph, wherein question and sentence intent recognition is carried out on the basis of received question information to obtain a semantic triple combination, question classification is carried out on the basis of the semantic triple combination according to a mapping rule to obtain a question category, semantic matching is carried out on the basis of the semantic triple combination and the question category to obtain a structured query language, matching query is carried out in a power failure knowledge graph established in advance on the basis of the structured query language to obtain an answer, matching judgment is carried out on the basis of the answer, and question-answering or knowledge graph updating and expansion are completed. Therefore, the power failure intelligent question-answering method based on the knowledge graph has the advantages of low question-answering dependency, accurate mapping and wide knowledge range, can greatly reduce the working pressure of power customer service, and meets the industry requirement of power technology informatization and intelligence.
Description
Technical Field
The application relates to the technical field of artificial intelligence, in particular to a power failure intelligent question answering method and device based on a knowledge graph.
Background
In 2019, the national generated energy is 75034.3 hundred million kilowatt hours, the year is 4.7 percent of the same proportion, in 2020, the social electricity consumption is 75110 million kilowatt hours, the year is 3.1 percent of the same proportion, and the safe operation of the power system has important significance for the development of the country and the society. For the problem in the fault treatment of the power system and the power equipment, the traditional solution is that a user or a power worker consults a power technology expert or searches technical data, and a large amount of time and manpower are consumed.
With the informatization and intelligent development of computer technology, the problem solving by using artificial intelligence technology becomes the trend of technical development, and various intelligent question-answering systems are also applied. The intelligent question-answering system is user-friendly and convenient to operate and use, can save time and energy spent on searching data or consulting technical experts, and can greatly reduce the pressure of technical customer service.
At present, no mature intelligent question-answering system exists for power failure. In a traditional question-answering system, a question-answering based on rule matching or a question-answering based on information retrieval is generally adopted; in the question and answer process based on rule matching, fuzzy matching is carried out by using the like keywords of the question and answer pair, so that the fact that the introduction of __, the description of __ and the __ of DTU have multiple reply modes can be found, and the precision is low; in the question and answer process based on information retrieval, it can be found that the word segmentation device of the search engine carries out word segmentation on input texts, such as 'DTU', 'power equipment', 'manufacturers' and 'common faults', and retrieval results are given by depending on comprehensive use of an inverted index table and a question template, and the mode seriously depends on manual construction of question and answer pairs, so that the recall rate is low. In addition, the contents of the database of the existing intelligent question-answering system mainly include the manufacturer of the equipment, the production date, the instruction book and the like, the knowledge range is very limited, and the questions posed by the user cannot be accurately understood.
Disclosure of Invention
The application provides an intelligent power failure question-answering method and device based on a knowledge graph, a query mode of combining question entity identification and question intention identification for mapping is adopted to solve the question-answering dependence problem and the problem of accurate mapping of a power inquiry mode and knowledge, and a power failure knowledge graph based on whole network technical data and updated in real time is arranged to solve the problem of limited knowledge range.
In a first aspect, the present application provides a power failure intelligent question-answering method based on a knowledge graph, including:
identifying question intention based on the received question information to obtain a semantic triple combination, wherein the semantic triple combination comprises one or more semantic triples;
performing problem classification according to a mapping rule based on the semantic triple combination to obtain a problem category;
performing semantic matching based on the semantic triple combination and the question category to obtain a structured query language;
matching query is carried out in a pre-established power failure knowledge graph based on the structured query language to obtain an answer;
performing matching judgment based on the answers, outputting the answers to finish question and answer if the answers are matched, and performing knowledge updating and expansion on the power failure knowledge graph and performing secondary query if the answers are not matched;
optionally, the identifying a question intention based on the received question information to obtain a semantic triple combination includes:
performing entity identification and entity attribute classification based on a deep learning model based on the problem information to obtain a preamble semantic triple combination;
performing similar statement normalization operation based on the preamble semantic triple combination to obtain the semantic triple combination;
optionally, the classifying the problem according to the mapping rule based on the semantic triple combination to obtain the problem category includes:
extracting the number of the semantic triples as first semantic information;
extracting the number of the semantic triples of which the relation p belongs to the common attribute expression set as second semantic information;
extracting the number of the semantic triples with the object o describing the state of the subject s as third semantic information;
extracting the number of the semantic triples with different subject words s as fourth semantic information;
judging whether the semantic triple combination meets the condition: the values of the first semantic information, the second semantic information and the third semantic information are all larger than 3, the value of the fourth semantic information is larger than 2, if the conditions are met, the problem is classified into a reasoning type, and if the conditions are not met, similarity calculation is carried out on the relation p + object o and the keyword sentence;
if the similarity reaches a threshold value, classifying the problem into a definition type, and if the similarity does not reach the threshold value, performing attribute judgment on the relation p;
and if the relation p belongs to the common attribute representation set, classifying the problem into an attribute label type, and if the relation p does not belong to the common attribute representation set, classifying the problem into an abnormal problem.
Optionally, the performing matching query in a pre-established power failure knowledge graph based on the structured query language to obtain an answer includes:
if the question is the definition type, the answer comprises definition content and attribute node content;
if the question is the attribute label type, the answer comprises the content related to the attribute node;
if the problem is the inference type, combining the semantics of the semantic triples, and carrying out knowledge inference by adopting an inference algorithm based on a knowledge graph path to generate an answer;
optionally, the method for establishing the power failure knowledge graph includes:
collecting technical data related to faults of power equipment and a power system;
carrying out entity identification and relation extraction based on the technical data to obtain a map semantic triple;
establishing the power failure knowledge graph based on the graph semantic triples;
optionally, the obtaining of the map semantic triple by performing entity identification and relationship extraction based on the technical data includes:
performing entity identification and relation extraction on the text of the technical data by using a deep learning model in a remote supervised learning mode to obtain the map semantic triple;
optionally, knowledge updating and expansion are performed on the established power failure knowledge graph periodically.
In a second aspect, the present application provides a power failure intelligent question-answering device based on a knowledge graph, the device comprising: the system comprises an interaction module, an intention identification module, a semantic matching module and a knowledge graph module;
the interaction module comprises: the device comprises an input module, an output module and a storage module;
the intent recognition module includes: the system comprises a semantic analysis module and a classification mapping module, wherein the semantic analysis module comprises an entity identification sub-module and an attribute classification sub-module, and the classification mapping module comprises a similarity normalization sub-module and a mapping rule sub-module;
the semantic matching module comprises: the query module and the answer module;
the knowledge-graph module comprises: the system comprises a data acquisition module, a knowledge extraction module, a knowledge storage module and a knowledge updating module;
the input module transmits the question information input by a user to the semantic analysis module, the entity recognition sub-module, the attribute classification sub-module and the similarity normalization sub-module perform question intention recognition and similar statement normalization operation on the question information to obtain the semantic triple combination, the attribute classification sub-module performs question classification based on a mapping rule to obtain the question category and transmits the semantic triple combination and the question category to the query module, the query module performs question shape judgment based on the semantic triple combination and the question category and performs semantic matching on the question with a normal shape to obtain the structured query language, the knowledge storage module performs query based on the structured query language to generate an answer, and the answer is subjected to matching judgment in the answer module, if the answers are matched, the answers are transmitted to the output module, if the answers are not matched, the question information and the answers are transmitted to the knowledge updating module to perform knowledge updating and expansion on the power failure knowledge graph, secondary query is performed, and the storage module stores the contents of the input module and the output module to form a user question-answer record;
the data acquisition module acquires technical data related to faults of power equipment and a power system, transmits the technical data to the knowledge extraction module, the knowledge extraction module performs entity identification and relation extraction based on the technical data to obtain map semantic triples, the power fault knowledge graph is established based on the map semantic triples, and the knowledge storage module stores the power fault knowledge graph established by the knowledge extraction module.
According to the technical scheme, the intelligent power failure question-answering method based on the knowledge graph is characterized in that question intention identification is carried out based on received question information to obtain a semantic triple combination, question classification is carried out according to mapping rules based on the semantic triple combination to obtain question categories, semantic matching is carried out based on the semantic triple combination and the question categories to obtain a structured query language, matching query is carried out in a pre-established power failure knowledge graph based on the structured query language to obtain answers, matching judgment is carried out based on the answers, the answers are output to complete question-answering if the answers are matched, knowledge updating and expanding are carried out on the power failure knowledge graph if the answers are not matched, and secondary query is carried out. The application provides an intelligent power failure question-answering method and device based on a knowledge graph, a query mode of mapping by combining question entity identification and question intention identification is adopted to solve the problem of question-answering dependency, a power inquiry mode and knowledge accurate mapping, and a power failure knowledge graph based on whole network technical data and updated in real time is arranged to solve the problem of limited knowledge range. According to the power failure intelligent question-answering method and device based on the knowledge graph, the working pressure of power customer service can be greatly reduced, the accuracy of answering questions by a traditional question-answering system is improved, and the industry requirement of power technology informatization and intellectualization is met.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without any creative effort.
FIG. 1 is a schematic diagram of a power failure intelligent question-answering method based on knowledge graph provided by the present application;
fig. 2 is a schematic diagram of a question and sentence intent recognition method provided in the present application;
FIG. 3 is a schematic diagram of a problem classification method provided herein;
FIG. 4 is a schematic diagram of a method for establishing a power failure knowledge graph according to the present application;
fig. 5 is a schematic structural diagram of an intelligent power failure question-answering device based on a knowledge graph.
Detailed Description
In a traditional power failure question-answering system, a question-answering based on rule matching or a question-answering based on information retrieval is generally adopted; the question and answer based on rule matching has various reply modes and is low in precision; the question and answer based on the information retrieval depends on comprehensive use of the inverted index table and the question template to give a retrieval result, and the question and answer pair is constructed by manpower seriously, so that the recall rate is low. In addition, the contents of the database of the question-answering system mainly include the manufacturer of the device, the date of manufacture, instructions for use, etc., the knowledge range is very limited, and the questions posed by the user cannot be accurately understood.
In order to solve the problems, the application provides an intelligent power failure question-answering method and device based on a knowledge graph, a question-answering dependency problem and a problem of accurate mapping of a power inquiry mode and knowledge are solved by adopting a query mode of mapping by combining question entity identification and question intention identification, and a problem of limited knowledge range is solved by setting a power failure knowledge graph based on full-network technical data and updated in real time.
Fig. 1 is a schematic diagram of a power failure intelligent question-answering method based on a knowledge graph according to an embodiment of the present application, and as shown in fig. 1, the method includes:
s1, identifying question and sentence intentions based on the received question information to obtain a semantic triple combination, wherein the semantic triple combination comprises one or more semantic triples;
further, as shown in fig. 2, in an embodiment of the present application, the performing question-and-sentence intent recognition based on the received question information to obtain a semantic triple combination includes:
s11, performing entity identification and entity attribute classification based on a deep learning model based on the problem information to obtain a preamble semantic triple combination; specifically, the natural language processing is carried out on the text of the question information, and the natural language processing comprises entity identification based on a Bi-LSTM + CRF deep learning model and entity attribute classification based on a CNN + Bi-GRU + FC deep learning model, wherein the question information is processed into the form of triples such as entity-relation-entity, entity-attribute-data and the like in a knowledge graph.
S12, carrying out similar statement normalization operation based on the preamble semantic triple combination to obtain the semantic triple combination; similar statement normalization is mainly to perform unified arrangement on the question information, since different users have different appellations for description of the same thing knowledge, but the Neo4j node information extracted and stored by the domain knowledge is standard power description of manual inspection, which causes the phenomenon that the same thing does not correspond to when inquiring, it is very necessary to perform statement arrangement on the knowledge of the question information in the process of question answering, and here, the similar statement normalization operation is performed on the basis of named entity identification and attribute supplement.
Compared with the traditional power failure question-answering system, the question-answering based on rule matching or the question-answering based on information retrieval is generally adopted, and the question-answering pairs are constructed manually by depending on the inverted index table and the question template, so that the precision and the recall rate are lower; the method and the device solve the problem of dependency of question answering and the problem of accurate mapping of electric power inquiry mode and knowledge by adopting a query mode of mapping combining question entity identification and question intention identification, and greatly improve the precision and recall rate.
S2, classifying the problems according to the mapping rules based on the semantic triple combination to obtain problem categories;
further, as shown in fig. 3, in an embodiment of the present application, the classifying the problem according to the mapping rule based on the semantic triple combination to obtain a problem category includes:
s21, extracting the number of the semantic triples as first semantic information;
s22, extracting the number of the semantic triples of which the relation p belongs to the common attribute expression set as second semantic information;
s23, extracting the number of the semantic triples with the object o describing the state of the subject S as third semantic information;
s24, extracting the number of the semantic triples with different subject words as fourth semantic information;
s25, judging whether the semantic triple combination meets the condition: the values of the first semantic information, the second semantic information and the third semantic information are all larger than 3, the value of the fourth semantic information is larger than 2, if the conditions are met, the problem is classified into a reasoning type, and if the conditions are not met, similarity calculation is carried out on the relation p + object o and the keyword sentence;
s26, if the similarity reaches a threshold value, classifying the problem into a definition type, and if the similarity does not reach the threshold value, performing attribute judgment on the relation p;
and S27, if the relation p belongs to the common attribute representation set, classifying the problem into an attribute label type, and if the relation p does not belong to the common attribute representation set, classifying the problem into an abnormal problem.
For example, the question information is: what is the power terminal equipment DTU? Identifying semantic triple combinations such as < DTU, power terminal equipment >, < DTU, and what > the semantic triple combinations, performing similar statement normalization operation, and classifying the semantic triple combinations into a definition type by combining with a mapping rule; for example, the question information is: manufacturer of DTU? Obtaining semantic triples such as < DTU, manufacturer > through question intention identification, and classifying the semantic triples into attribute label types if the manufacturers belong to an attribute set; for example, the question information is: the method comprises the steps that a signal lamp of the DTU equipment displays a red light, the equipment stops running, fault phenomenon description languages such as a screen blue screen of the equipment and the like are identified through question and sentence intention, and semantic triple combinations are obtained, such as a plurality of semantic triples including a signal lamp > and a DTU including a screen, > < signal lamp, state, red light >, < equipment, running state, stop >, < screen, display, blue screen and the like, and object o in the semantic triples is mainly described as a state of a subject s, and then the semantic triples are classified into reasoning types.
S3, performing semantic matching based on the semantic triple combination and the question category to obtain a structured query language;
s4, carrying out matching query in a pre-established power failure knowledge graph based on the structured query language to obtain an answer;
further, in this embodiment of the present application, the performing a matching query in a pre-established power failure knowledge graph based on the structured query language to obtain an answer includes:
if the question is the definition type, the answer comprises definition content and attribute node content;
if the question is the attribute label type, the answer comprises the content related to the attribute node;
and if the problem is the inference type, combining the semantics of the semantic triples, and carrying out knowledge inference by adopting an inference algorithm based on a knowledge graph path to generate an answer.
For example, if the question information is a definition type, the attribute node is searched as the supplementary information to be returned to form an answer while the definition content is returned; for example, if the question information is of an attribute tag type, only information related to the attribute node is returned; for example, if the question information is inference type, combining semantics of a plurality of triples is required, and performing a knowledge inference to obtain an answer by using an inference algorithm based on a knowledge graph path, where the question information is: relation of DTU device to FTU device? The semantic triples such as < DTU, relation and FTU > are identified through question intentions and converted into structured query languages (Match p ═ (n1: entry 1) - [ r: rel ] - > (n2: entry 2) where ne n1.name ═ 0} 'and n2.name ═ 1}' return distin. rel. where name1 and name2 are entity names and rel is equivalent to the relation between two entities), FTU nodes are searched from DTU nodes in the knowledge graph along the path during query, if the FTU nodes cannot be found, no relation is returned, and if the FTU nodes are found successfully, the relation path and the intermediate nodes are returned as combined answers.
And S5, performing matching judgment based on the answers, outputting the answers to finish question answering if the answers are matched, and performing knowledge updating and expansion on the power failure knowledge graph and performing secondary query if the answers are not matched.
Further, as shown in fig. 4, in an embodiment of the present application, the method of establishing the power failure knowledge-graph before performing the power failure intelligent question-answering method based on the knowledge-graph includes:
s01, collecting technical data related to faults of the power equipment and the power system;
s02, performing entity identification and relation extraction based on the technical data to obtain a map semantic triple;
and S03, establishing the power failure knowledge graph based on the graph semantic triples.
Further, in an embodiment of the present application, the obtaining of the atlas semantic triple through entity identification and relationship extraction based on the technical data includes:
and carrying out entity identification and relation extraction on the text of the technical data by using a deep learning model in a remote supervised learning mode to obtain the map semantic triple.
Further, in the embodiment of the present application, knowledge updating and expansion are performed on the established power failure knowledge map periodically.
The knowledge graph is mainly used as a resource library of the intelligent question-answering system, the knowledge graph can contain large-scale and abundant semantic information, and compared with a traditional character string matching retrieval mode, the semantic matching retrieval mode based on the knowledge graph is beneficial to accurately acquiring the problem intention and providing matched answers. Technical data related to faults of the power equipment and the power system are collected by using a crawler technology, technical documents related to faults of the power equipment and the power system on a network and books including technical posts of an authoritative power technology forum are legally obtained; taking related technical documents such as a power equipment manual, a common power system fault solution and the like as model training corpora, and performing entity identification and relation extraction on a technical text by using a deep learning model (Bert + Bi-LSTM + CRF) in a remote supervised learning mode to obtain a map semantic triple; constructing a power equipment fault knowledge graph by using graph semantic triples extracted from texts, and storing the knowledge graph by using a non-relational database Neo4 j; and regularly updating and expanding knowledge of the constructed knowledge graph.
The non-relational database Neo4j is a high-performance NoSQL graph database that stores data as nodes and relationships between nodes in a graph. Different entities in Neo4j are associated through various relationships to form a complex object graph, and the functions of searching and traversing on the object graph, namely, deep search and breadth search, are realized. The embodiment adopts the non-relational database Neo4j to store the knowledge graph, which is beneficial to improving the retrieval precision and recall rate.
Fig. 5 is a schematic structural diagram of a power failure intelligent question-answering device based on a knowledge graph according to an embodiment of the present application, and as shown in fig. 5, the device includes: interaction module 100, intent recognition module 200, semantic matching module 300, and knowledge graph module 400;
the interaction module 100 includes: an input module 110, an output module 120, and a storage module 130;
the intent recognition module 200 includes: the system comprises a semantic analysis module 210 and a classification mapping module 220, wherein the semantic analysis module 210 comprises an entity identification submodule 211 and an attribute classification submodule 212, and the classification mapping module 220 comprises a similarity normalization submodule 221 and a mapping rule submodule 222;
the semantic matching module 300 includes: a query module 310 and an answer module 320;
the knowledge-graph module 400 includes: the system comprises a data acquisition module 410, a knowledge extraction module 420, a knowledge storage module 430 and a knowledge updating module 440;
the input module 110 transmits the question information input by the user to the semantic analysis module 210, the entity identification sub-module 211 and the attribute classification sub-module 212 perform question intent identification on the question information to obtain a preamble semantic triple combination, and transmit the preamble semantic triple combination to the classification mapping module 220, the similarity normalization sub-module 221 performs similar statement normalization operation to obtain the semantic triple combination, the attribute classification sub-module 212 performs question classification based on a mapping rule to obtain the question category, and transmits the semantic combination and the question category to the query module 310, the query module 310 performs question shape judgment based on the semantic triple combination and the question category, and performs semantic matching on the question with a normal question shape to obtain the structured query language, querying is performed on the knowledge storage module 430 based on the structured query language to generate an answer, matching judgment is performed on the answer in an answer module 320, if the answer is matched, the answer is transmitted to the output module 120, if the answer is not matched, the question information and the answer are transmitted to the knowledge updating module 440 to perform knowledge updating and expansion on the power failure knowledge graph, secondary query is performed, and the storage module 130 stores the contents of the input module 110 and the output module 120 to form a user question-answer record; by storing and recording the question and answer records, the user can conveniently inquire historical question and answer, repeated question and answer is reduced, and the working efficiency is improved.
The data acquisition module 410 acquires technical data about faults of power equipment and a power system, transmits the technical data to the knowledge extraction module 420, the knowledge extraction module 420 performs entity identification and relationship extraction based on the technical data to obtain map semantic triples, establishes the power fault knowledge map based on the map semantic triples, and the knowledge storage module 430 stores the power fault knowledge map established by the knowledge extraction module 420. The knowledge storage module 430 stores the knowledge graph by using the non-relational database Neo4j, and realizes the functions of searching and traversing on the object graph, namely, depth search and breadth search. The embodiment adopts the non-relational database Neo4j to store the knowledge graph, which is beneficial to improving the retrieval precision and recall rate.
The apparatus provided in the embodiment of the present application may be configured to execute any of the question answering methods shown in the above embodiments, and the implementation manner and the technical effect of the apparatus are similar to each other.
According to the technical scheme, the power failure intelligent question-answering method and device based on the knowledge graph are provided, the question-answering problem of dependency and the problem of accurate mapping of a power inquiry mode and knowledge are solved by adopting an inquiry mode of combining question entity identification and question intention identification for mapping, and the problem of limited knowledge range is solved by setting the power failure knowledge graph based on the whole network technical data and updated in real time. According to the power failure intelligent question-answering method and device based on the knowledge graph, the working pressure of power customer service can be greatly reduced, the accuracy of answering questions by a traditional question-answering system is improved, and the industry requirement of power technology informatization and intellectualization is met.
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the disclosed invention. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
Claims (8)
1. A power failure intelligent question-answering method based on a knowledge graph is characterized by comprising the following steps:
identifying question intention based on the received question information to obtain a semantic triple combination, wherein the semantic triple combination comprises one or more semantic triples;
performing problem classification according to a mapping rule based on the semantic triple combination to obtain a problem category;
performing semantic matching based on the semantic triple combination and the question category to obtain a structured query language;
matching query is carried out in a pre-established power failure knowledge graph based on the structured query language to obtain an answer;
and performing matching judgment based on the answers, outputting the answers to finish question and answer if the answers are matched, and performing knowledge updating and expansion on the power failure knowledge graph and performing secondary query if the answers are not matched.
2. The method of claim 1, wherein the performing question intent recognition based on the received question information results in a semantic triple combination comprising:
performing entity identification and entity attribute classification based on a deep learning model based on the problem information to obtain a preamble semantic triple combination;
and carrying out similar statement normalization operation based on the preamble semantic triple combination to obtain the semantic triple combination.
3. The method of claim 1, wherein the problem classification based on the semantic triple combination according to a mapping rule to obtain a problem category comprises:
extracting the number of the semantic triples as first semantic information;
extracting the number of the semantic triples of which the relation p belongs to the common attribute expression set as second semantic information;
extracting the number of the semantic triples with the object o describing the state of the subject s as third semantic information;
extracting the number of the semantic triples with different subject words s as fourth semantic information;
judging whether the semantic triple combination meets the condition: the values of the first semantic information, the second semantic information and the third semantic information are all larger than 3, and the value of the fourth semantic information is larger than 2; if the condition is met, classifying the problem into an inference type, and if the condition is not met, calculating the similarity of the relation p + object o and the keyword sentence;
if the similarity reaches a threshold value, classifying the problem into a definition type, and if the similarity does not reach the threshold value, performing attribute judgment on the relation p;
and if the relation p belongs to the common attribute representation set, classifying the problem into an attribute label type, and if the relation p does not belong to the common attribute representation set, classifying the problem into an abnormal problem.
4. The method of claim 3, wherein said matching queries in a pre-established power failure knowledge graph based on said structured query language to obtain answers comprises:
if the question is the definition type, the answer comprises definition content and attribute node content;
if the question is the attribute label type, the answer comprises the content related to the attribute node;
and if the problem is the inference type, combining the semantics of the semantic triples, and carrying out knowledge inference by adopting an inference algorithm based on a knowledge graph path to generate an answer.
5. The method of claim 1, wherein the power failure knowledge map is created by a method comprising:
collecting technical data related to faults of power equipment and a power system;
carrying out entity identification and relation extraction based on the technical data to obtain a map semantic triple;
and establishing the power failure knowledge graph based on the graph semantic triples.
6. The method of claim 5, wherein the performing entity identification and relationship extraction based on the technical material to obtain graph-semantic triples comprises:
and carrying out entity identification and relation extraction on the text of the technical data by using a deep learning model in a remote supervised learning mode to obtain the map semantic triple.
7. The method of claim 6, further comprising periodically updating and augmenting knowledge of the established power failure knowledge map.
8. A knowledge-graph-based power failure intelligent question-answering device applied to the method of any one of claims 1 to 7, wherein the device comprises: an interaction module (100), an intent recognition module (200), a semantic matching module (300), and a knowledge-graph module (400);
the interaction module (100) comprises: an input module (110), an output module (120) and a storage module (130);
the intent recognition module (200) comprises: a semantic analysis module (210) and a classification mapping module (220), the semantic analysis module (210) comprising an entity identification sub-module (211) and an attribute classification sub-module (212), the classification mapping module (220) comprising a similarity normalization sub-module (221) and a mapping rule sub-module (222);
the semantic matching module (300) comprises: a query module (310) and an answer module (320);
the knowledge-graph module (400) comprises: the system comprises a data acquisition module (410), a knowledge extraction module (420), a knowledge storage module (430) and a knowledge updating module (440);
the input module (110) transmits the question information input by the user to the semantic analysis module (210), the entity identification sub-module (211), the attribute classification sub-module (212) and the similarity normalization sub-module (221) perform question intention identification and similar statement normalization on the question information to obtain the semantic triple combination, the attribute classification sub-module (212) performs question classification based on mapping rules to obtain the question category, and transmits the semantic triple combination and the question category to the query module (310), the query module (310) performs question shape judgment based on the semantic triple combination and the question category, performs semantic matching on the question with a normal question shape to obtain the structured query language, and performs query in the knowledge storage module (430) based on the structured query language to generate answers, matching and judging the answers in an answer module (320), if the answers are matched, transmitting the answers to the output module (120), if the answers are not matched, transmitting the question information and the answers to a knowledge updating module (440), performing knowledge updating and expansion on the power failure knowledge graph, and performing secondary query, wherein the storage module (130) stores the contents of the input module (110) and the output module (120) to form a user question-answer record;
the data acquisition module (410) acquires technical data related to faults of power equipment and a power system, the technical data are transmitted to the knowledge extraction module (420), the knowledge extraction module (420) performs entity identification and relation extraction based on the technical data to obtain map semantic triples, the power fault knowledge map is established based on the map semantic triples, and the knowledge storage module (430) stores the power fault knowledge map established by the knowledge extraction module (420).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110377904.6A CN113051382A (en) | 2021-04-08 | 2021-04-08 | Intelligent power failure question-answering method and device based on knowledge graph |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110377904.6A CN113051382A (en) | 2021-04-08 | 2021-04-08 | Intelligent power failure question-answering method and device based on knowledge graph |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113051382A true CN113051382A (en) | 2021-06-29 |
Family
ID=76519260
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110377904.6A Pending CN113051382A (en) | 2021-04-08 | 2021-04-08 | Intelligent power failure question-answering method and device based on knowledge graph |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113051382A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113535449A (en) * | 2021-07-08 | 2021-10-22 | 平安科技(深圳)有限公司 | Abnormal event repairing processing method and device, computer equipment and storage medium |
CN113609268A (en) * | 2021-07-30 | 2021-11-05 | 徐涛 | Intelligent psychological knowledge question-answering method and device based on knowledge graph |
CN113672720A (en) * | 2021-09-14 | 2021-11-19 | 国网天津市电力公司 | Power audit question and answer method based on knowledge graph and semantic similarity |
CN115309870A (en) * | 2022-10-11 | 2022-11-08 | 启元世界(北京)信息技术服务有限公司 | Knowledge acquisition method and device |
CN115374108A (en) * | 2022-07-22 | 2022-11-22 | 北京三维天地科技股份有限公司 | Data standard generation and automatic mapping method based on knowledge graph technology |
CN116821304A (en) * | 2023-07-07 | 2023-09-29 | 国网青海省电力公司信息通信公司 | Knowledge intelligent question-answering system of power supply station based on big data |
CN117290484A (en) * | 2023-10-10 | 2023-12-26 | 深圳市明睿数据科技有限公司 | Intelligent question-answering system and method for automobile fault positioning and maintenance suggestion |
CN118211171A (en) * | 2024-05-22 | 2024-06-18 | 沈阳安华晟源信息科技有限公司 | Knowledge graph-based target path mining method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109062939A (en) * | 2018-06-20 | 2018-12-21 | 广东外语外贸大学 | A kind of intelligence towards Chinese international education leads method |
CN109767081A (en) * | 2018-12-25 | 2019-05-17 | 云南电网有限责任公司信息中心 | Electric power work order table generating method and device |
CN110717018A (en) * | 2019-04-15 | 2020-01-21 | 中国石油大学(华东) | Industrial equipment fault maintenance question-answering system based on knowledge graph |
CN110765277A (en) * | 2019-10-22 | 2020-02-07 | 河海大学常州校区 | Online equipment fault diagnosis platform of mobile terminal based on knowledge graph |
CN112199488A (en) * | 2020-11-04 | 2021-01-08 | 国网江苏省电力有限公司营销服务中心 | Incremental knowledge graph entity extraction method and system for power customer service question answering |
-
2021
- 2021-04-08 CN CN202110377904.6A patent/CN113051382A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109062939A (en) * | 2018-06-20 | 2018-12-21 | 广东外语外贸大学 | A kind of intelligence towards Chinese international education leads method |
CN109767081A (en) * | 2018-12-25 | 2019-05-17 | 云南电网有限责任公司信息中心 | Electric power work order table generating method and device |
CN110717018A (en) * | 2019-04-15 | 2020-01-21 | 中国石油大学(华东) | Industrial equipment fault maintenance question-answering system based on knowledge graph |
CN110765277A (en) * | 2019-10-22 | 2020-02-07 | 河海大学常州校区 | Online equipment fault diagnosis platform of mobile terminal based on knowledge graph |
CN112199488A (en) * | 2020-11-04 | 2021-01-08 | 国网江苏省电力有限公司营销服务中心 | Incremental knowledge graph entity extraction method and system for power customer service question answering |
Non-Patent Citations (4)
Title |
---|
刘瑞宏 等: "基于知识图谱的智能故障诊断研究", 《邮电设计技术》 * |
张敏杰 等: "面向变压器智能运检的知识图谱构建和智能问答技术研究", 《全球能源互联网》 * |
曾帅 等: "面向知识自动化的自动问答研究进展", 《自动化学报》 * |
邓显波 等: "基于知识图谱的电缆故障诊断分析", 《现代信息科技》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113535449A (en) * | 2021-07-08 | 2021-10-22 | 平安科技(深圳)有限公司 | Abnormal event repairing processing method and device, computer equipment and storage medium |
CN113535449B (en) * | 2021-07-08 | 2024-05-03 | 平安科技(深圳)有限公司 | Abnormal event restoration processing method and device, computer equipment and storage medium |
CN113609268A (en) * | 2021-07-30 | 2021-11-05 | 徐涛 | Intelligent psychological knowledge question-answering method and device based on knowledge graph |
CN113672720A (en) * | 2021-09-14 | 2021-11-19 | 国网天津市电力公司 | Power audit question and answer method based on knowledge graph and semantic similarity |
CN115374108A (en) * | 2022-07-22 | 2022-11-22 | 北京三维天地科技股份有限公司 | Data standard generation and automatic mapping method based on knowledge graph technology |
CN115374108B (en) * | 2022-07-22 | 2023-06-20 | 北京三维天地科技股份有限公司 | Knowledge graph technology-based data standard generation and automatic mapping method |
CN115309870A (en) * | 2022-10-11 | 2022-11-08 | 启元世界(北京)信息技术服务有限公司 | Knowledge acquisition method and device |
CN116821304A (en) * | 2023-07-07 | 2023-09-29 | 国网青海省电力公司信息通信公司 | Knowledge intelligent question-answering system of power supply station based on big data |
CN116821304B (en) * | 2023-07-07 | 2023-12-19 | 国网青海省电力公司信息通信公司 | Knowledge intelligent question-answering system of power supply station based on big data |
CN117290484A (en) * | 2023-10-10 | 2023-12-26 | 深圳市明睿数据科技有限公司 | Intelligent question-answering system and method for automobile fault positioning and maintenance suggestion |
CN118211171A (en) * | 2024-05-22 | 2024-06-18 | 沈阳安华晟源信息科技有限公司 | Knowledge graph-based target path mining method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113051382A (en) | Intelligent power failure question-answering method and device based on knowledge graph | |
CN109284363B (en) | Question answering method and device, electronic equipment and storage medium | |
CN102087669B (en) | Intelligent search engine system based on semantic association | |
CN110765277B (en) | Knowledge-graph-based mobile terminal online equipment fault diagnosis method | |
CN114911951A (en) | Knowledge graph construction method for man-machine cooperation assembly task | |
CN112507076A (en) | Semantic analysis searching method and device and storage medium | |
CN115757810A (en) | Method for constructing standard ontology of knowledge graph | |
US11487795B2 (en) | Template-based automatic software bug question and answer method | |
CN113626571A (en) | Answer sentence generating method and device, computer equipment and storage medium | |
CN117909754A (en) | Auxiliary power plant equipment defect elimination method and system based on twin neural network | |
CN117668182A (en) | Standard intelligent question-answering method and system integrating knowledge graph and large language model | |
CN115937881A (en) | Method for automatically identifying content of knowledge graph construction standard form | |
CN115964466A (en) | Intelligent question-answering method and system based on four-layer feature vector matching model | |
CN114691878A (en) | Construction method of automobile standard knowledge graph | |
CN113792157A (en) | Domain mechanism-oriented knowledge base construction method | |
Dai et al. | QAM: question answering system based on knowledge graph in the military | |
Lin et al. | Research and application of knowledge graph technology for intelligent question answering | |
CN117874240B (en) | Audit text classification method, system and equipment based on knowledge graph | |
Yan et al. | Implementation of Intelligent Q&A System for Electric Power Knowledge Based on Knowledge Graph | |
CN114722163B (en) | Data query method and device, electronic equipment and storage medium | |
CN115248854B (en) | Automatic question-answering method, system and storage medium based on knowledge graph | |
Xu et al. | Research on Knowledge Graph-Based Q&A for Aero-Engine Fault Diagnosis | |
CN118069716B (en) | Auxiliary decision making system based on knowledge enhancement strong model in group type enterprise background | |
Feng et al. | Intelligent question answering system based on entrepreneurial incubation knowledge graph | |
Deng et al. | An Artificial Intelligence Model Recommendation Method for Power Dispatching Scenario Based on Knowledge Graph and Scene Label Matching |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210629 |
|
RJ01 | Rejection of invention patent application after publication |