CN113010663A - Adaptive reasoning question-answering method and system based on industrial cognitive map - Google Patents
Adaptive reasoning question-answering method and system based on industrial cognitive map Download PDFInfo
- Publication number
- CN113010663A CN113010663A CN202110463562.XA CN202110463562A CN113010663A CN 113010663 A CN113010663 A CN 113010663A CN 202110463562 A CN202110463562 A CN 202110463562A CN 113010663 A CN113010663 A CN 113010663A
- Authority
- CN
- China
- Prior art keywords
- industrial
- question
- query
- cognitive
- map
- 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
- 230000001149 cognitive effect Effects 0.000 title claims abstract description 71
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000003044 adaptive effect Effects 0.000 title claims description 18
- 230000009977 dual effect Effects 0.000 claims abstract description 11
- 230000004927 fusion Effects 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 13
- 238000010276 construction Methods 0.000 claims description 11
- 238000005516 engineering process Methods 0.000 claims description 11
- 238000004519 manufacturing process Methods 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 238000003672 processing method Methods 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims 1
- 230000003068 static effect Effects 0.000 abstract description 5
- 230000008447 perception Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000019771 cognition Effects 0.000 description 2
- 230000003920 cognitive function Effects 0.000 description 2
- 238000001303 quality assessment method Methods 0.000 description 2
- 238000013441 quality evaluation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000009434 installation 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—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
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Human Computer Interaction (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Manufacturing & Machinery (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a self-use reasoning question-answering method and a self-use reasoning question-answering system based on an industrial cognitive map, which comprise the following steps: acquiring an industrial natural language query question input by industrial related personnel; extracting entities in the query sentence and the relation thereof to match with the industrial cognitive map to obtain a local link map; obtaining an answer to the query question according to the matched knowledge graph, or performing multi-hop reasoning by using a perception-cognition dual system to obtain the answer to the query question; matching the query question answers with the question output model and outputting the question answers. By adopting the technical scheme of the invention, the fusion between static characteristic data and dynamic time sequence data in the current industrial scene is solved, and the adaptability of industrial natural language query and the accuracy rate of question answering are improved.
Description
Technical Field
The application belongs to the technical field of cognitive inference, and particularly relates to a self-adaptive inference question-answering method and system based on an industrial cognitive atlas.
Background
With the rise of artificial intelligence application, big data grows in an immeasurable explosion form, the development of artificial intelligence technology is promoted, people face the problem that data cannot be accurately acquired while acquiring numerous convenient information, therefore, the effective search engine technology is very important for the use of people in specific fields, the search engine is essentially used for collecting information from the Internet by using a computer according to a certain strategy, and then after the information is processed and organized, search service is provided for users, and the information is displayed in a manner convenient for the users to understand.
In the industrial field, a large amount of semi-structured data and unstructured data exist, wherein the semi-structured data mostly refers to storage data in the format of XML, JSON and the like in the industry, the unstructured data mostly refers to process documents, pictures, texts, process videos, audio information and the like in an industrial scene, and the data are required to be processed into structured data for the next operation. For large-scale different types of data, it is important to convert the data into usable knowledge by extracting and utilizing the data. Meanwhile, in an industrial scene, besides the static data, a large amount of processing dynamic data also exists, the data is important for mastering various conditions of the whole life cycle of manufacturing, and how to sense, acquire and store the dynamic data is to be researched.
Disclosure of Invention
The invention aims to solve the technical problem of providing a self-use reasoning question-answering method and system based on an industrial cognitive map, solving the problem of fusion between static characteristic data and dynamic time sequence data in the current industrial scene, and improving the adaptability of industrial natural language query and the question-answering accuracy.
In order to achieve the purpose, the invention adopts the following technical scheme:
an adaptive reasoning question-answering system based on an industrial cognitive map comprises:
acquiring an industrial natural language query question input by industrial related personnel;
extracting entities in the query sentence and the relation thereof to match with the industrial cognitive map to obtain a local link map;
obtaining an answer to the query question according to the matched knowledge graph, or performing multi-hop reasoning by using a perception-cognition dual system to obtain the answer to the query question;
matching the query question answers with the question output model and outputting the question answers.
Preferably, the processing technology involved in the process manufacturing process is subjected to knowledge extraction, knowledge fusion and knowledge processing methods to construct the industrial cognitive map.
Preferably, the industrial natural language query question contains at least one entity and at least one relationship between the relationships.
Preferably, a local link map containing query relations and entities is formed according to key entities and relations extracted from the industrial natural language query question, the problem type is judged according to the scale of the local link map and the industrial natural language query question, and whether the problem is a one-hop problem or a multi-hop problem is judged.
Preferably, if the query question of the industrial natural language is a one-hop question, directly obtaining an answer of the query question according to the matched local link map; and matching the initial industrial cognitive map by using the industrial query statement candidate query data set, searching the entity name and the relationship name, and selecting the entity and the entity at the other end of the relationship link as answers if the query statement only comprises one entity and one group of relationships.
Preferably, if the industrial natural language query question is a multi-hop problem, a perception-cognition dual system is used for multi-hop reasoning to obtain a query question answer; it includes:
step 201, arranging multiple types of sensors at different positions in a factory to sense different types of data, and acquiring dynamic time sequence data of a product in a full life cycle;
step 202, extracting key knowledge and converting the entities and the relations in the industrial query problem, other clues of the industrial query statement and other process documents related to the entities into semantic vectors through BERT; label prediction and classification are carried out on the embedded characters or words by using the LSTM; adding constraint to the predicted label by using CRF;
step 203, implicit reasoning calculation is realized through GNN, each iteration step transfers the transformed information to the next reasoning through a preamble node, and updates the current implicit representation, and the implicit representation of the answer candidate node judges which is the final answer through a full-connection network with a softmax function.
Preferably, all entities and relations in the industrial query statement are extracted and stored in a specified candidate query data set;
searching and matching all entities and relations in the candidate data set in the initial industrial cognitive map in sequence, if the entities and relations in the data set cannot be completely established with link relations, selecting a certain entity as a main node to match the rest nodes and relations, and requiring the obtained local link map to contain all the entities and relations in the industrial query statement;
performing first-step query according to an industrial query statement, retrieving a known entity and a known relation to obtain a required answer, and taking the adjacent relation of nodes corresponding to the entity and the known relation as a center to obtain a semantic relation of the next step;
judging whether the obtained next-order semantic relationship has a query result of the industrial query statement, if so, retaining the result, and if not, searching the next-order semantic relationship again, wherein the next-order semantic relationship diagram is a next-order entity for extracting the entity and a semantic link relationship thereof, and while obtaining the next-order semantic relationship, obtaining similar semantics in a candidate query data set corresponding to the semantic relationship, updating an initial cognitive map, and forming a cyclic cognitive map;
and reasoning by using the updated cognitive map until a question-answer result of the industrial query statement is obtained.
The invention also provides an adaptive reasoning question-answering system based on the industrial cognitive map, which comprises the following components:
the acquisition module is used for acquiring an industrial natural language query sentence input by industrial related personnel;
the extraction module is used for extracting the entity and the relation in the query sentence to be matched with the industrial cognitive map to obtain a local link map;
the reasoning module is used for obtaining an answer of the query question according to the matched knowledge graph or carrying out multi-hop reasoning by utilizing a perception-cognition dual system to obtain the answer of the query question;
and the matching module is used for matching the query question answers with the question output model and outputting the question answers.
Preferably, the method further comprises the following steps: and the construction module is used for constructing the industrial cognitive map by the processing technology involved in the process manufacturing process through knowledge extraction, knowledge fusion and knowledge processing methods.
The invention relates to a self-adaptive reasoning question-answering method and a self-adaptive reasoning question-answering system based on an industrial cognitive map, wherein the knowledge map can convert data with simple significance into useful knowledge, the knowledge map is essentially a semantic network, the semantic network technology provides a query environment for a user, the core of the query environment is to return processed and inferred knowledge to the user in a graphic mode, and the knowledge map technology is a foundation and a bridge for realizing intelligent semantic retrieval. The cognitive map is the fusion of the knowledge map with cognitive science, cognitive computation and the like, and the cognitive map realizes self-circulation, self-updating and self-adaptation on the basis of the knowledge map, so that the quality and the reasoning function of the cognitive map are favorable for improving the inquiry, question and answer of workers to various problems in the whole life cycle of product manufacturing aiming at an industrial system with complex data types. By adopting the technical scheme of the invention, the fusion between static characteristic data and dynamic time sequence data in the current industrial scene is solved, and the adaptability of industrial natural language query and the accuracy rate of question answering are improved.
Drawings
Fig. 1 is a schematic flow chart of an adaptive inference question-answering method based on an industrial cognitive atlas according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a construction process of an industrial cognitive map;
FIG. 3 is a diagram of reasoning for only one entity and only one relationship between entities in an industrial natural language query question according to an embodiment of the present invention;
fig. 4 is a diagram of reasoning for various entities and relationships contained in an industrial natural language query question according to the embodiment of the present invention;
FIG. 5 is a schematic diagram of a cognitive-cognitive dual system provided by an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of an adaptive inference question-answering system based on an industrial knowledge graph according to an embodiment of the present invention;
FIG. 7 is a flow of reasoning based on one-hop question-and-answer industrial query statements provided by an embodiment of the present invention;
fig. 8 is an inference flow based on multi-hop question-and-answer industrial query statements provided in an embodiment of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiment of the present invention will be clearly and completely described below with reference to the drawings in the embodiment of the present invention, and it is obvious that the described embodiment is only a part of the embodiment of the present invention, and not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
A schematic flow diagram of an adaptive inference question-answering method based on an industrial cognitive map provided in an embodiment of the present specification is shown in fig. 1, and the method includes:
step 100, constructing an industrial cognitive map
And constructing the initial industrial cognitive map by processing technologies involved in the process manufacturing process through methods of knowledge extraction, knowledge fusion, knowledge processing and the like.
The embodiment of the invention provides a flow chart for building an industrial cognitive map, and as shown in fig. 2, the method may include:
step 1001, entity extraction
Structured, semi-structured, unstructured data in an industrial scenario primarily includes process documents, operation manuals, video feeds, etc., which are generally unorganized and processed discrete, objective ground truth, with industrial entities being abstracted through different methods of named entity identification.
Step 1002, entity linking and knowledge merging
The data are cleaned and integrated through entity linking and knowledge merging, ambiguity is eliminated, redundancy and wrong concepts are eliminated, the quality of the data is guaranteed, the entity linking is mainly realized through industrial entity objects extracted from process texts and linked to corresponding correct industrial entity objects in an industrial knowledge base, and then large-scale knowledge merging is carried out through a knowledge merging method when an external industrial scene knowledge base or existing structured industrial knowledge input is received.
Step 1003, ontology construction, knowledge reasoning and quality evaluation
The extracted entities are not equal to knowledge, so ontology construction, knowledge reasoning and quality assessment are needed to form the industrial knowledge into a structured and networked knowledge system. The ontology construction is mainly a specification for modeling knowledge in an industrial scene, and definition is given to industrial concepts and relations thereof in a formalized manner. Knowledge inference refers to extracting new associations from existing industrial entity relationship data through inference of logic or graphs, so that the whole industrial knowledge graph is enriched. Quality assessment refers to the quality of the industrial knowledge base.
And then, inputting an industrial query statement by operators of all stations in the manufacturing scene, requiring the query statement to at least comprise one entity and at least one relation between the entities, and extracting key entities and relations from the obtained industrial query statement.
Step 101, obtaining an industrial natural language query question input by industrial related personnel
Because the types of industry related people are more when the industry related knowledge is queried, the industry natural language query question sentence can at least comprise at least one entity and at least one relationship between the entities and the relationships.
Step 102, extracting entities in query sentences and matching the entities and the relations thereof with the industrial cognitive map to obtain a local link map
The method comprises the steps that an industrial natural language query question input by industrial related personnel is used, wherein entities with parts of speech such as nouns and verbs are involved, and the extracted entities are stored in a candidate query data set through named entity recognition.
And forming a local link map containing query relations and entities according to key entities and relations extracted from the industrial natural language query question, judging the problem type according to the scale of the local link map and the industrial natural language query question, and judging whether the problem is a one-hop problem or a multi-hop problem.
103, if the query question of the industrial natural language is a one-hop question, directly obtaining an answer of the query question according to the matched local link map
Matching of the initial industrial cognitive map is performed by using the candidate query data set of the industrial query statement, the entity name and the relationship name are searched, if the query statement only contains one entity and one group of relationships, the entity and the entity at the other end of the relationship link are selected as answers, and the scale of the local link map generally contains one group of entities, as shown in fig. 3.
In general, an entity and a relationship may correspond to a plurality of tail entities, and therefore, it is necessary to perform manual selection and judgment among the tail entities to determine the answer to the question, as shown in fig. 7.
For example, "what tool is needed for turning? The industrial query statement identifies the entity and the relation of the statement by naming entity identification, and identifies that the entity has turning and tools, wherein the tools are used as labels of answers and are not used as key entities, and the relation is identified as using;
the turning and the using are respectively used as head nodes and relations to be searched and matched in the industrial cognitive map, a plurality of tail entity nodes of 75-degree external turning tools, 45-degree end face turning tools, threading tools and inner bore turning tools are obtained through traversal, at the moment, industrial relevant personnel independently judge answers through a series of provided information, and judge results as the 45-degree external turning tools according to the reduced answers.
Step 104, if the query sentence of the industrial natural language is a multi-hop problem, the perception-cognition dual system is used for multi-hop reasoning to obtain the answer of the query sentence
Generally, an industrial query statement needs a clear natural language question-answer mode, if a plurality of entities and a plurality of groups of relations are involved, one-step searching cannot be realized to obtain answers generally, and therefore, multi-step analysis is needed to obtain answers;
extracting all entities and relations in the industrial query statement and storing the entities and relations in a specified candidate query data set;
searching and matching all entities and relations in the candidate data set in the initial industrial cognitive map in sequence, if the entities and relations in the data set cannot be completely established with link relations, selecting a certain entity as a main node to match the rest nodes and relations, and requiring the obtained local link map to contain all the entities and relations in the industrial query statement;
performing first-step query according to an industrial query statement, retrieving a known entity and a known relation to obtain a required answer, and taking the adjacent relation of nodes corresponding to the entity and the known relation as a center to obtain a semantic relation of the next step;
judging whether the obtained next-order semantic relationship has a query result of the industrial query statement, if so, retaining the result, and if not, searching the next-order semantic relationship again, wherein the next-order semantic relationship diagram is a next-order entity for extracting the entity and a semantic link relationship thereof, and while obtaining the next-order semantic relationship, obtaining similar semantics in a candidate query data set corresponding to the semantic relationship, updating an initial cognitive map, and forming a cyclic cognitive map;
and reasoning by using the updated cognitive map until a question-answer result of the industrial query statement is obtained.
Generally, an industrial natural language query question does not simply relate to a one-hop problem, and for a case that a plurality of entities and relations exist in the question, answer query is performed in a multi-hop problem search mode, as shown in fig. 4.
The required answer cannot be directly obtained through the first-step query, so that the next-order semantic relationship of the adjacent relationship of the corresponding nodes of the entity needs to be obtained, and if the required answer cannot be obtained yet, the next node and relationship need to be searched until the relevant content is searched, as shown in fig. 8.
For example, "what equipment is needed for turning an end face? The industrial query statement identifies the entity and the relation of the statement by naming entity identification, and identifies that the entity has turning, an end surface and equipment, wherein the equipment is used as a label of an answer and is not used as a key entity, and the identified relation is processing;
the two entity nodes of turning and end face and the processing relation are searched and matched in the industrial cognitive map, and are just combined into a group of triples of turning, processing and end face, but the answer label is 'equipment', so that the next hop of search is needed.
And (3) combining key information of the 'end face', taking 'turning' as a head entity node, and matching to a '45-degree end face lathe tool', thereby obtaining a next triple { turning, using and 45-degree end face lathe tool }, wherein the answer label is 'equipment', so that the search of the next hop is required.
The 45-degree end face turning tool is used as an entity and has a correlation with other two entities, wherein one entity is used as a previous search answer, and the other correlation is a lathe, so that the triple of the jump reasoning is { 45-degree end face turning tool, installation and lathe }.
The classification of the lathe has more nodes, and at the moment, the answers are already narrowed to a range as small as possible, and industrial users only need to search for the answers within the range.
The embodiment of the present invention provides a perception-cognition dual system for assisting answer reasoning of a multi-hop question, and requires to continuously update iteration to obtain a question-answer result, as shown in fig. 5, the method may include:
step 201, obtaining dynamic time series data through a sensor
The system 1 (perception system) performs a perception function, namely updating of dynamic time series data in the system
Firstly, acquiring and processing dynamic time-sequence data of a product in a full life cycle, and mainly sensing different types of data by arranging various types of sensors at different positions in a factory.
Because the time sequence data has the sequence characteristics, different time step lengths are set for different sensing data, and then the data are acquired in different nodes. The dynamic data is generally parameter type data with time stamp, and is converted into a uniform format to be stored in a database.
Step 202, vector embedding is carried out on the industrial natural language query sentence
The system 1 executes the perception function of updating the related knowledge nodes involved in the question-answering sentences
Constructing entity and relation extraction training corpora, and respectively training an entity and a relation extraction model, wherein the extraction model adopts a BERT + LSTM + CRF model;
extracting key knowledge and converting the entities and the relations in the industrial query problem, other clues of the industrial query statement and other process documents related to the entities into semantic vectors through BERT;
then, label prediction and classification are carried out on the embedded characters or words by using the LSTM;
and finally, adding constraints on the predicted labels by adopting the CRF so as to ensure that the predicted labels are legal. And these extracted and partitioned knowledge will be stored in different databases.
Step 203, the system 2 (authentication system) executes the cognitive function, and a series of implicit reasoning is realized through GNN
And (3) realizing implicit reasoning calculation through GNN, transmitting the transformed information to the next reasoning through preamble nodes in each iteration step, updating the current implicit representation, and judging which is the final answer through a full-connection network with a softmax function by the implicit representation of the answer candidate nodes.
And then, obtaining the industrial query statement result aiming at the relationship graph obtained by various query reasoning, and sending the relevant question and answer corpus to the user according to the way which can be understood by the user.
Step 105, matching the query question answer with the question output model to output the question answer
The result obtained through the self-adaptive inference of the cognitive map is generally output as simple word class representation, and the result is replaced to the position of a designated noun aiming at a simple industrial query statement;
the example shows for a simple word class that the question and answer "what tool is needed for turning? And the step of replacing according to the searched result is that turning needs to use a 45-degree external turning tool.
Aiming at the complex industrial query sentences, training answer output models are carried out on sample data of existing question-answering sentences in the cognitive map, and allelic replacement is carried out on complex question-answering results.
An embodiment of the present specification provides an adaptive inference question-answering system based on an industrial cognitive atlas, which implements the above adaptive inference question-answering method, and as shown in fig. 6, the question-answering system provided by the embodiment of the present specification includes:
the acquisition module is used for acquiring an industrial natural language query sentence input by industrial related personnel;
the extraction module is used for extracting the entity and the relation in the query sentence to be matched with the industrial cognitive map to obtain a local link map;
the reasoning module is used for obtaining an answer of the query question according to the matched knowledge graph or carrying out multi-hop reasoning by utilizing a perception-cognition dual system to obtain the answer of the query question;
and the matching module is used for matching the query question answers with the question output model and outputting the question answers.
Further, still include: and the construction module is used for constructing the industrial cognitive map by the processing technology involved in the process manufacturing process through knowledge extraction, knowledge fusion and knowledge processing methods.
The self-adaptive reasoning question-answering method and system based on the industrial cognitive map realize the construction of the initial industrial cognitive map by utilizing the related process of the construction of the knowledge map aiming at a large amount of heterogeneous, multidimensional and time sequence data in an industrial scene. Firstly, extracting entities, relations and attributes according to a large amount of heterogeneous data in an industrial scene, wherein the results may contain a large amount of redundant and wrong information, so that the quality of knowledge is ensured through knowledge fusion, and then ontology construction, knowledge reasoning, quality evaluation and the like are realized through knowledge processing. The method comprises the steps of utilizing relevant technologies such as cognitive science and the like to realize updating iteration of a cognitive map and cognitive reasoning and the like to realize knowledge question answering aiming at processing natural language problems, adopting a dual-system mode to realize reasoning question answering, and respectively carrying out sensing and cognition based on the cognitive map, wherein the sensing function of a system 1 realizes the construction of an initial map and the knowledge updating function based on question answering sentences, the initial cognitive map comprises the respective sensing of static data and dynamic data, the cognitive function of a system 2 realizes the reasoning based on the cognitive map, and a neural network is adopted to carry out the cycle of reasoning results, so that the reasoning of implicit relations and results is realized, and the knowledge question answering is carried out. The result is continuously updated with the cognitive map of the system 1, and the whole cycle alternation from perception to cognition is realized. The invention also discloses a self-adaptive reasoning question-answering system based on the industrial cognitive map; the auxiliary factory design and manufacturing personnel realize the accuracy and rapidity of searching.
Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.
Claims (9)
1. An adaptive reasoning question-answering system based on an industrial cognitive map is characterized by comprising:
acquiring an industrial natural language query question input by industrial related personnel;
extracting entities in the query sentence and the relation thereof to match with the industrial cognitive map to obtain a local link map;
obtaining an answer to the query question according to the matched knowledge graph, or performing multi-hop reasoning by using a perception-cognition dual system to obtain the answer to the query question;
matching the query question answers with the question output model and outputting the question answers.
2. The adaptive inference question-answering system based on industrial cognitive maps of claim 1, further comprising: and constructing the industrial cognitive map by processing technologies involved in the process manufacturing process through knowledge extraction, knowledge fusion and knowledge processing methods.
3. The adaptive inference question-answering system based on industrial cognitive maps according to claim 1 or 2, wherein the industrial natural language query question contains at least one entity and at least one relationship between the relationships.
4. The adaptive inference question-answering system based on industrial cognitive mapping of claim 3, wherein a local link map including query relations and entities is formed according to key entities and relations extracted from the industrial natural language query question, and a question type is judged according to the scale of the local link map and the industrial natural language query question to judge whether the question is a one-hop question or a multi-hop question.
5. The adaptive inference question-answering system based on the industrial cognitive atlas of claim 4, wherein if the industrial natural language query question is a one-hop question, the answer of the query question is directly obtained according to the matched local link atlas; and matching the initial industrial cognitive map by using the industrial query statement candidate query data set, searching the entity name and the relationship name, and selecting the entity and the entity at the other end of the relationship link as answers if the query statement only comprises one entity and one group of relationships.
6. The adaptive inference question-answering system based on the industrial cognitive atlas as claimed in claim 4, wherein if the industrial natural language query question is a multi-hop problem, a perception-cognition dual system is used for multi-hop inference to obtain an answer to the query question; it includes:
step 201, arranging multiple types of sensors at different positions in a factory to sense different types of data, and acquiring dynamic time sequence data of a product in a full life cycle;
step 202, extracting key knowledge and converting the entities and the relations in the industrial query problem, other clues of the industrial query statement and other process documents related to the entities into semantic vectors through BERT; label prediction and classification are carried out on the embedded characters or words by using the LSTM; adding constraint to the predicted label by using CRF;
step 203, implicit reasoning calculation is realized through GNN, each iteration step transfers the transformed information to the next reasoning through a preamble node, and updates the current implicit representation, and the implicit representation of the answer candidate node judges which is the final answer through a full-connection network with a softmax function.
7. The adaptive inference question-answering system based on industrial cognitive profiles of claim 1, wherein,
extracting all entities and relations in the industrial query statement and storing the entities and relations in a specified candidate query data set;
searching and matching all entities and relations in the candidate data set in the initial industrial cognitive map in sequence, if the entities and relations in the data set cannot establish all link relations, selecting a certain entity as a main node to match the rest nodes and relations, and requiring the obtained local link map to contain all the entities and relations in the industrial query statement;
performing first-step query according to an industrial query statement, retrieving a known entity and a known relation to obtain a required answer, and taking the adjacent relation of nodes corresponding to the entity and the known relation as a center to obtain a semantic relation of the next step;
judging whether the obtained next-order semantic relationship has a query result of the industrial query statement, if so, retaining the result, and if not, searching the next-order semantic relationship again, wherein the next-order semantic relationship diagram is a next-order entity for extracting the entity and a semantic link relationship thereof, and while obtaining the next-order semantic relationship, obtaining similar semantics in a candidate query data set corresponding to the semantic relationship, updating an initial cognitive map, and forming a cyclic cognitive map;
and reasoning by using the updated cognitive map until a question-answer result of the industrial query statement is obtained.
8. An adaptive reasoning question-answering system based on an industrial cognitive map is characterized by comprising:
the acquisition module is used for acquiring an industrial natural language query sentence input by industrial related personnel;
the extraction module is used for extracting the entity and the relation in the query sentence to be matched with the industrial cognitive map to obtain a local link map;
the reasoning module is used for obtaining an answer of the query question according to the matched knowledge graph or carrying out multi-hop reasoning by utilizing a perception-cognition dual system to obtain the answer of the query question;
and the matching module is used for matching the query question answers with the question output model and outputting the question answers.
9. The adaptive inference question-answering system based on industrial cognitive maps of claim 8, further comprising: and the construction module is used for constructing the industrial cognitive map by the processing technology involved in the process manufacturing process through knowledge extraction, knowledge fusion and knowledge processing methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110463562.XA CN113010663A (en) | 2021-04-26 | 2021-04-26 | Adaptive reasoning question-answering method and system based on industrial cognitive map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110463562.XA CN113010663A (en) | 2021-04-26 | 2021-04-26 | Adaptive reasoning question-answering method and system based on industrial cognitive map |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113010663A true CN113010663A (en) | 2021-06-22 |
Family
ID=76380721
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110463562.XA Pending CN113010663A (en) | 2021-04-26 | 2021-04-26 | Adaptive reasoning question-answering method and system based on industrial cognitive map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113010663A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113779220A (en) * | 2021-09-13 | 2021-12-10 | 内蒙古工业大学 | Mongolian multi-hop question-answering method based on three-channel cognitive map and graph attention network |
| CN114399051A (en) * | 2021-12-29 | 2022-04-26 | 北方工业大学 | Intelligent food safety question-answer reasoning method and device |
| CN114496234A (en) * | 2022-04-18 | 2022-05-13 | 浙江大学 | Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients |
| CN115248854A (en) * | 2022-09-22 | 2022-10-28 | 中邮消费金融有限公司 | Knowledge graph-based automatic question answering method, system and storage medium |
| CN115964459A (en) * | 2021-12-28 | 2023-04-14 | 北方工业大学 | Multi-hop inference question-answering method and system based on food safety cognitive map |
| CN116340530A (en) * | 2023-02-17 | 2023-06-27 | 江苏科技大学 | Intelligent design method based on mechanical knowledge graph |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107679124A (en) * | 2017-09-21 | 2018-02-09 | 北京航空航天大学 | A kind of knowledge mapping Chinese question and answer search method based on dynamic programming algorithm |
| CN109033135A (en) * | 2018-06-06 | 2018-12-18 | 北京大学 | A kind of natural language querying method and system of software-oriented project knowledge map |
| CN110704600A (en) * | 2019-09-30 | 2020-01-17 | 北京百度网讯科技有限公司 | Question-answer dynamic matching method and device and electronic equipment |
| CN110837550A (en) * | 2019-11-11 | 2020-02-25 | 中山大学 | Knowledge graph-based question and answer method and device, electronic equipment and storage medium |
| CN111209412A (en) * | 2020-02-10 | 2020-05-29 | 同方知网(北京)技术有限公司 | Method for building knowledge graph of periodical literature by cyclic updating iteration |
| CN112148851A (en) * | 2020-09-09 | 2020-12-29 | 常州大学 | Construction method of medicine knowledge question-answering system based on knowledge graph |
| CN112527997A (en) * | 2020-12-18 | 2021-03-19 | 中国南方电网有限责任公司 | Intelligent question-answering method and system based on power grid field scheduling scene knowledge graph |
-
2021
- 2021-04-26 CN CN202110463562.XA patent/CN113010663A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107679124A (en) * | 2017-09-21 | 2018-02-09 | 北京航空航天大学 | A kind of knowledge mapping Chinese question and answer search method based on dynamic programming algorithm |
| CN109033135A (en) * | 2018-06-06 | 2018-12-18 | 北京大学 | A kind of natural language querying method and system of software-oriented project knowledge map |
| CN110704600A (en) * | 2019-09-30 | 2020-01-17 | 北京百度网讯科技有限公司 | Question-answer dynamic matching method and device and electronic equipment |
| CN110837550A (en) * | 2019-11-11 | 2020-02-25 | 中山大学 | Knowledge graph-based question and answer method and device, electronic equipment and storage medium |
| CN111209412A (en) * | 2020-02-10 | 2020-05-29 | 同方知网(北京)技术有限公司 | Method for building knowledge graph of periodical literature by cyclic updating iteration |
| CN112148851A (en) * | 2020-09-09 | 2020-12-29 | 常州大学 | Construction method of medicine knowledge question-answering system based on knowledge graph |
| CN112527997A (en) * | 2020-12-18 | 2021-03-19 | 中国南方电网有限责任公司 | Intelligent question-answering method and system based on power grid field scheduling scene knowledge graph |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113779220A (en) * | 2021-09-13 | 2021-12-10 | 内蒙古工业大学 | Mongolian multi-hop question-answering method based on three-channel cognitive map and graph attention network |
| CN115964459A (en) * | 2021-12-28 | 2023-04-14 | 北方工业大学 | Multi-hop inference question-answering method and system based on food safety cognitive map |
| CN115964459B (en) * | 2021-12-28 | 2023-09-12 | 北方工业大学 | Multi-hop reasoning question-answering method and system based on food safety cognition spectrum |
| CN114399051A (en) * | 2021-12-29 | 2022-04-26 | 北方工业大学 | Intelligent food safety question-answer reasoning method and device |
| CN114496234A (en) * | 2022-04-18 | 2022-05-13 | 浙江大学 | Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients |
| CN115248854A (en) * | 2022-09-22 | 2022-10-28 | 中邮消费金融有限公司 | Knowledge graph-based automatic question answering method, system and storage medium |
| CN115248854B (en) * | 2022-09-22 | 2023-03-03 | 中邮消费金融有限公司 | Automatic question-answering method, system and storage medium based on knowledge graph |
| CN116340530A (en) * | 2023-02-17 | 2023-06-27 | 江苏科技大学 | Intelligent design method based on mechanical knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113010663A (en) | Adaptive reasoning question-answering method and system based on industrial cognitive map | |
| CN110990590A (en) | Dynamic financial knowledge map construction method based on reinforcement learning and transfer learning | |
| CN101901247B (en) | Vertical engine searching method and system for domain body restraint | |
| CN112507136B (en) | Knowledge-driven business operation map construction method | |
| CN108932342A (en) | A kind of method of semantic matches, the learning method of model and server | |
| CN111651447B (en) | Intelligent construction life-span data processing, analyzing and controlling system | |
| KR20050062624A (en) | Learning/thinking machine and learning/thinking method based on structured knowledge, computer system, and information generation method | |
| Prudhomme et al. | Interpretation and automatic integration of geospatial data into the Semantic Web: Towards a process of automatic geospatial data interpretation, classification and integration using semantic technologies | |
| CN110245238A (en) | The figure embedding grammar and system of Process Based and syntax schema | |
| CN111339269A (en) | Knowledge graph question-answer training and application service system with automatically generated template | |
| CN112948546A (en) | Intelligent question and answer method and device for multi-source heterogeneous data source | |
| EP3864564A1 (en) | System for searching natural language documents | |
| CN112163097A (en) | Military knowledge graph construction method and system | |
| CN117151338B (en) | Multi-unmanned aerial vehicle task planning method based on large language model | |
| KR20200082179A (en) | Data transformation method for spatial data's semantic annotation | |
| CN110909124B (en) | Hybrid enhanced intelligent demand accurate sensing method and system based on human-in-loop | |
| CN117271767A (en) | Operation and maintenance knowledge base establishing method based on multiple intelligent agents | |
| CN116861269A (en) | Multi-source heterogeneous data fusion and analysis method in engineering field | |
| CN117312578B (en) | Construction method and system of non-genetic carrier spectrum | |
| CN108959366B (en) | Open question-answering method | |
| CN114168615A (en) | Method and system for querying SCD (substation configuration description) file of intelligent substation by natural language | |
| CN117094390A (en) | Knowledge graph construction and intelligent search method oriented to ocean engineering field | |
| Bharambe et al. | Ontology and knowledge graphs for semantic analysis in natural language processing | |
| CN113468311B (en) | Knowledge graph-based complex question and answer method, device and storage medium | |
| Sheremet | Data and Knowledge Bases with Incomplete Information in a Set of Strings Framework |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210622 |