A system for determining right and wrong based on a clause library
By using a clause-based discrimination system, which incorporates a thesaurus module, clause library, standard question-and-answer module, combined question-and-answer module, and discrimination module, the problem of quickly and accurately judging clauses in the field of electrical engineering is solved, achieving intelligent and rapid judgment and efficient clause retrieval.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- STATE GRID ANHUI ELECTRIC POWER CO LTD ELECTRIC POWER SCI RES INST
- Filing Date
- 2022-06-07
- Publication Date
- 2026-05-12
AI Technical Summary
Newcomers to the field of electrical engineering or those outside the field often find it difficult to quickly and accurately identify the relevant clauses in electrical problems, leading to low search efficiency and hindering industry development.
The judgment system based on the clause library includes a thesaurus module, a clause library, a standard question and answer module, a combined question and answer module, an association module, and a judgment module. By establishing a topic thesaurus, a standard question and answer library, a combined question and answer library, and similarity calculation, it can quickly make right and wrong judgments.
It enables intelligent and rapid identification of user issues, reduces invalid searches, improves work efficiency, helps users quickly obtain the corresponding terms, and promotes industry development.
Smart Images
Figure CN115048498B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electrical knowledge question-and-answer technology, specifically a yes / no discrimination system based on a clause library. Background Technology
[0002] With the rapid development of the electrical field, related professional knowledge and standards are constantly being updated, which has brought some trouble to the work of relevant personnel. In particular, for those who are new to the field or those who are not in the field, it takes a lot of effort to find out whether an electrical problem is correct and what the corresponding clause is, which is not conducive to the development of the industry. Therefore, this invention provides a right-or-wrong judgment system based on a clause library to help users judge whether the corresponding problem is correct. Summary of the Invention
[0003] To address the problems of the above solutions, this invention provides a yes / no discrimination system based on a clause library.
[0004] The objective of this invention can be achieved through the following technical solutions:
[0005] The yes / no discrimination system based on the clause library includes a thesaurus module, a clause library, a standard question and answer module, a combined question and answer module, an association module, a discrimination module, and a server;
[0006] The thesaurus module is used to establish a thesaurus; the clause library stores corresponding standard clauses; the standard question-and-answer module establishes a corresponding standard question-and-answer library based on the standard clauses in the clause library; the combined question-and-answer module is used to establish a combined question-and-answer library based on the thesaurus, establish a statement model, combine the thesaurus terms in the thesaurus using the established statement model, output combined questions and answers, output the thesaurus term set of the combined questions and answers, perform numerical conversion of the thesaurus term set to obtain the thesaurus term set value; set the context value of the corresponding combined questions and answers, integrate the context value and the corresponding thesaurus term set value to obtain a matching vector, map the matching vector to a vector space, and summarize the current combined questions and answers to establish a combined question-and-answer library;
[0007] The association module links combined questions and answers with standard questions and answers; the discrimination module is used to determine whether the identified question is correct, obtain the identified user question, convert the identified user question into a question vector, input the question vector into the corresponding vector space, calculate the similarity between the question vector and the corresponding matching vector, match the corresponding combined questions and answers according to the calculated similarity, obtain the corresponding standard questions and answers according to the matched combined questions and answers, match the corresponding standard clauses according to the obtained standard questions and answers, and make a right or wrong judgment on the user question according to the standard questions and answers and standard clauses to obtain the corresponding judgment result.
[0008] Furthermore, the working method of the thesaurus module includes:
[0009] Set up a dictionary and define rules, perform format conversion for review, and obtain standard constraints; establish a database and data collection channels, collect data through the established channels, and input the collected data into the database for storage. Mark the database as a material library, extract keywords from the data in the material library, mark them as initial words, verify the initial words through the set standard constraints, mark the verified initial words as subject words, integrate the subject words, and establish a subject word library.
[0010] Furthermore, the working method of the standard question-and-answer module includes:
[0011] Identify the standard clauses in the clause library, obtain the application direction of the standard clauses, set the corresponding question search formula based on the obtained application direction, perform question search according to the set question search formula, obtain a number of question and answer questions, remove duplicates from the obtained question and answer questions, obtain candidate question and answer questions, filter the candidate question and answer questions to obtain standard question and answer questions, and label them with the corresponding standard clause tags, and integrate the obtained standard question and answer questions to build a standard question and answer library.
[0012] Furthermore, methods for setting contextual values for corresponding question-and-answer combinations include:
[0013] Identify the corresponding keyword set values, labeled ZTi, where i represents a keyword, i = 1, 2, ..., n, and n is a positive integer; identify the standard question and answer corresponding to the combined question and answer, match the corresponding standard additional value, labeled BZ; set the context adjustment coefficient for the combined question and answer, labeled α, according to the context value formula. Calculate the context value, where b1 and b2 are both proportionality coefficients, with a value range of 0. <b1≤1,0<b2≤1。
[0014] Furthermore, the working method of the associated module includes:
[0015] Identify combined questions and answers in the combined question and answer database and standard questions and answers in the standard question and answer database. Calculate the similarity between the combined questions and answers and the standard questions and answers to obtain the corresponding similarity scores. Based on the similarity scores, assign the combined questions and answers to the corresponding standard questions and answers, assign the corresponding subordinate tags, and establish a subordinate question and answer table.
[0016] Furthermore, methods for transforming identified user questions into question vectors include:
[0017] Extract keywords from user questions to obtain a keyword set, assign values to the keyword set, set context values for user questions, and integrate them into a corresponding question vector.
[0018] Furthermore, methods for assigning values to the obtained set of question keywords include:
[0019] Match the set of question keywords with the keywords in the thesaurus to obtain the corresponding values. Mark the question keywords that do not match as similar keywords. Match the similar keywords with the corresponding similar values and integrate them into the keyword set.
[0020] Furthermore, the method for assigning similarity values based on matching similar keywords is as follows:
[0021] Similar keywords are input into a thesaurus for semantic similarity calculation. The highest similarity score is obtained, along with the corresponding topic word's value, which is labeled XS and FZ respectively. The formula is then used to calculate the similarity. Calculate the corresponding similarity assignment, where b3 is the scaling factor, with a value ranging from 0 to 1. <b3≤1, This is the similarity adjustment factor.
[0022] Compared with the prior art, the beneficial effects of the present invention are:
[0023] By coordinating the thesaurus module, clause library, standard Q&A module, combined Q&A module, association module, and discrimination module, the system enables rapid yes / no judgment when a user enters a question, and simultaneously retrieves the corresponding clauses. When the user needs the clauses, the system can retrieve them, achieving intelligent judgment of the question. This greatly assists users in answering questions, avoids invalid searches, and provides support for the development of the industry and enterprises. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 This is a block diagram illustrating the principle of the present invention. Detailed Implementation
[0026] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0027] like Figure 1 As shown, the yes / no discrimination system based on the clause library includes a thesaurus module, a clause library, a standard question and answer module, a combined question and answer module, an association module, a discrimination module, and a server;
[0028] The thesaurus module is used to build a topic thesaurus, and the specific methods include:
[0029] Set up a dictionary and define rules, perform format conversion for review, and obtain standard constraints; establish a database and data collection channels, collect data through the established channels, and input the collected data into the database for storage. Mark the database as a material library, extract keywords from the data in the material library, mark them as initial words, verify the initial words through the set standard constraints, mark the verified initial words as subject words, integrate the subject words, and establish a subject word library.
[0030] Material data that has been used in the database will not be used again.
[0031] The process of extracting keywords from the data in the resource library is common knowledge in this field, and therefore will not be described in detail.
[0032] The initial words are checked by setting standard constraints. The expert group can build a corresponding learning model and check them by using the established learning model. The checked keywords can also be checked again by manual review as needed.
[0033] The dictionary's rules are set manually, with the main principles being that keywords should be concise and clear, extracted from power grid equipment technical standards, or derived from power-related terms relevant to the technical standards in the entire process of power grid equipment management. These terms must belong to the power industry and be reviewed and approved after being determined through business operations following frequency statistics and standardization processing. The dictionary should also include the following constraints:
[0034] The thesaurus should be drawn from a large number of domestic and foreign power grid equipment technical documents and archives, and selected after frequency statistics and standardization. The thesaurus should be based on concepts, and can include single words and compound words, but must be single-meaning words with clear concepts and strong specificity. The thesaurus should not be too long and should be easy to input and search for by computers. The thesaurus generally includes terms in five aspects: entity, behavior, state, time, and space. The selected thesaurus should be nouns or noun phrases. The selected thesaurus should be compatible with relevant thesaurus lists at home and abroad as much as possible.
[0035] The process of converting the review format involves transforming manually defined dictionary rules into system-recognized language rules, thereby obtaining standard constraints that can be directly recognized and judged by the system. This conversion can be carried out using existing technologies.
[0036] The data collection channels are used to collect dictionary text data. They are set up manually, first determining the corresponding dictionary coverage area, and then establishing corresponding data collection channels based on that coverage area. Data collection within the specified dictionary coverage area can be achieved using existing technical solutions. The dictionary coverage area includes power-related terms and jargon from the main and secondary standards of 72 types of power grid equipment, such as transformers, circuit breakers, and voltage transformers. Reference can be made to professional terms involved in power grid equipment management, such as power-related terms and jargon consistent with technical standards used in the entire management process of power grid equipment planning feasibility studies, engineering design, equipment procurement, equipment manufacturing, equipment acceptance, equipment installation, equipment commissioning, final acceptance, operation and maintenance, and decommissioning.
[0037] The term library is used to store corresponding standard terms to provide standard data support for question and answer judgment. The specific term library can use existing ones, or be supplemented or rebuilt manually.
[0038] The standard question and answer module is used to build a corresponding standard question and answer library based on the standard clauses in the clause library. Specific methods include:
[0039] Identify the standard clauses in the clause library, obtain the application direction of the standard clauses, set the corresponding question search formula based on the obtained application direction, perform question search according to the set question search formula, obtain a number of question and answer questions, remove duplicates from the obtained question and answer questions, obtain candidate question and answer questions, filter the candidate question and answer questions to obtain standard question and answer questions, and label them with the corresponding standard clause tags, and integrate the obtained standard question and answer questions to build a standard question and answer library.
[0040] The question search terms set based on the obtained application directions are common knowledge in the field, so they will not be described in detail; the candidate questions and answers are screened manually; the specific parts not disclosed in this application are common technical knowledge in the field, so they will not be described in detail.
[0041] The combined question-and-answer module is used to build a combined question-and-answer library based on a topic thesaurus, establish a statement model, combine topic terms in the topic thesaurus using the established statement model, output combined questions and answers, output a topic term set for the combined questions and answers, perform numerical conversion of the topic term set to obtain the topic term set value; set the context value for the corresponding combined questions and answers, integrate the context value and the corresponding topic term set value to obtain a matching vector, map the matching vector to a vector space, and summarize the current combined questions and answers to build a combined question-and-answer library.
[0042] Methods for setting context values for corresponding question-and-answer combinations include:
[0043] Identify the corresponding keyword set values, labeled ZTi, where i represents a keyword, i = 1, 2, ..., n, and n is a positive integer; identify the standard question and answer corresponding to the combined question and answer, match the corresponding standard additional value, labeled BZ; set the context adjustment coefficient for the combined question and answer, labeled α, according to the context value formula. Calculate the context value, where b1 and b2 are both proportionality coefficients, with a value range of 0. <b1≤1,0<b2≤1。
[0044] By setting context values, the matching accuracy of the matching vector in the subsequent question-and-answer matching process can be improved, thereby improving the accuracy of the yes / no judgment.
[0045] The system identifies the standard questions and answers corresponding to the combined questions and answers. The corresponding standard questions and answers can be obtained through the association module. A corresponding standard additional value is set for each standard question and answer. The expert group sets the standard additional value and establishes a corresponding standard additional value matching table. After matching, the corresponding standard additional value is obtained.
[0046] The method for numerical conversion of the thesaurus is as follows: establish a corresponding assignment matching table based on the existing thesaurus, perform matching based on the assignment matching table, and obtain the corresponding thesaurus assignment.
[0047] The sentence model is built based on a CNN or DNN network and then trained by setting a corresponding training set. The specific building and training process is common knowledge in this field.
[0048] Setting the context adjustment coefficient for combined question and answer can be done by simultaneously outputting the corresponding context adjustment coefficient when building and training the sentence model, or by building a new neural network model and setting it.
[0049] The association module is used to associate combined questions and answers with standard questions and answers, identify combined questions and answers in the combined question and answer library and standard questions and answers in the standard question and answer library, calculate the similarity between combined questions and answers and standard questions and answers to obtain the corresponding similarity, assign combined questions and answers to corresponding standard questions and answers according to the similarity, assign corresponding subordinate tags, and establish a subordinate question and answer table.
[0050] The similarity between the combined question-and-answer and the standard question-and-answer is calculated directly using existing similarity algorithms.
[0051] Assigning combined questions and answers to corresponding standard questions and answers based on similarity means identifying combined questions and answers and standard questions and answers with corresponding meanings based on the calculated similarity. This is because the number of combined questions and answers is generally higher than that of standard questions and answers; that is, generally speaking, one standard question and answer may correspond to multiple combined questions and answers.
[0052] The discrimination module is used to determine whether the identified question is correct, and the specific methods include:
[0053] The process involves acquiring identified user questions, extracting keywords from these questions to obtain a keyword set, assigning values to the keyword set, setting contextual values for the user questions, integrating these into a corresponding question vector, inputting the question vector into the corresponding vector space, calculating the similarity between the question vector and the corresponding matching vector, matching corresponding question-and-answer combinations based on the calculated similarity, obtaining corresponding standard questions and answers based on the matched question-and-answer combinations, matching corresponding standard clauses based on the obtained standard questions and answers, and judging the user questions as true or false based on the standard questions and answers and standard clauses to obtain the corresponding judgment result.
[0054] Methods for assigning values to the obtained set of question keywords include:
[0055] Match the set of question keywords with the keywords in the thesaurus to obtain the corresponding values. Mark the question keywords that do not match as similar keywords. Match the similar keywords with the corresponding similar values and integrate them into the keyword set.
[0056] The method for assigning similarity values based on similar keywords is as follows:
[0057] Similar keywords are input into a thesaurus for semantic similarity calculation. The highest similarity score is obtained, along with the corresponding thesaurus values, which are labeled XS and FZ respectively. These values are distinguished from the previous thesaurus values using a different representation method, based on the formula. Calculate the corresponding similarity assignment, where b3 is the scaling factor, with a value ranging from 0 to 1. <b3≤1, The similarity adjustment factor was set through discussion by the expert panel.
[0058] Similar keywords are input into a thesaurus for semantic similarity calculation. This can be done by manually setting up a corresponding learning model for intelligent calculation, or by building a similar thesaurus for different themes and classifying the similar words into different levels with different similarities. After matching, the corresponding similarity is obtained. The specific undisclosed parts are common knowledge in this field.
[0059] The context value of the user's question is set in a way that is different from the context value of the combined question and answer. The corresponding neural network model is directly built and trained. The intelligent setting is achieved by using artificial intelligence. Although there is some error, the error is within the allowable range. The specific building and training process is common knowledge in this field.
[0060] Judging the truth or falsehood of user questions based on standard questions and answers and standard terms can be achieved through existing technologies or by establishing corresponding learning models. This is common knowledge in the field, so it will not be described in detail.
[0061] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.
[0062] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A no-fault judgment system based on a clause library, characterized in that, It includes a thesaurus module, an item library, a standard question and answer module, a combined question and answer module, an association module, a discrimination module, and a server; The thesaurus module is used to establish a thesaurus; the clause library stores corresponding standard clauses; the standard question-and-answer module establishes a corresponding standard question-and-answer library based on the standard clauses in the clause library; the combined question-and-answer module is used to establish a combined question-and-answer library based on the thesaurus, establish a statement model, combine the thesaurus terms in the thesaurus using the established statement model, output combined questions and answers, output the thesaurus term set of the combined questions and answers, perform numerical conversion of the thesaurus term set to obtain the thesaurus term set value; set the context value of the corresponding combined questions and answers, integrate the context value and the corresponding thesaurus term set value to obtain a matching vector, map the matching vector to a vector space, and summarize the current combined questions and answers to establish a combined question-and-answer library; The association module associates combined questions and answers with standard questions and answers; the discrimination module is used to determine whether the identified question is correct, obtain the identified user question, convert the identified user question into a question vector, input the question vector into the corresponding vector space, calculate the similarity between the question vector and the corresponding matching vector, match the corresponding combined questions and answers according to the calculated similarity, obtain the corresponding standard questions and answers according to the matched combined questions and answers, match the corresponding standard clauses according to the obtained standard questions and answers, and make a right or wrong judgment on the user question according to the standard questions and answers and standard clauses to obtain the corresponding judgment result. Methods for setting context values for corresponding question-and-answer combinations include: Identify the corresponding keyword set values, labeled ZTi, where i represents a keyword, i = 1, 2, ..., n, and n is a positive integer; identify the standard question and answer corresponding to the combined question and answer, match the corresponding standard additional value, labeled BZ; set the context adjustment coefficient for the combined question and answer, labeled α, according to the context value formula. Calculate the context value, where b1 and b2 are both proportionality coefficients, with a value range of 0. <b1≤1,0<b2≤1。 2. The yes / no discrimination system based on a clause library according to claim 1, characterized in that, The working methods of the thesaurus module include: Set up a dictionary and define rules, perform format conversion for review, and obtain standard constraints; establish a database and data collection channels, collect data through the established channels, and input the collected data into the database for storage. Mark the database as a material library, extract keywords from the data in the material library, mark them as initial words, verify the initial words through the set standard constraints, mark the verified initial words as subject words, integrate the subject words, and establish a subject word library.
3. The yes / no discrimination system based on a clause library according to claim 1, characterized in that, The standard question-and-answer module works by including: Identify the standard clauses in the clause library, obtain the application direction of the standard clauses, set the corresponding question search formula based on the obtained application direction, perform question search according to the set question search formula, obtain a number of question and answer questions, remove duplicates from the obtained question and answer questions, obtain candidate question and answer questions, filter the candidate question and answer questions to obtain standard question and answer questions, and label them with the corresponding standard clause tags, and integrate the obtained standard question and answer questions to build a standard question and answer library.
4. The yes / no discrimination system based on a clause library according to claim 1, characterized in that, The working methods of the associated modules include: Identify combined questions and answers in the combined question and answer database and standard questions and answers in the standard question and answer database. Calculate the similarity between the combined questions and answers and the standard questions and answers to obtain the corresponding similarity scores. Based on the similarity scores, assign the combined questions and answers to the corresponding standard questions and answers, assign the corresponding subordinate tags, and establish a subordinate question and answer table.
5. The yes / no discrimination system based on a clause library according to claim 1, characterized in that, Methods for converting identified user questions into question vectors include: Extract keywords from user questions to obtain a keyword set, assign values to the keyword set, set context values for user questions, and integrate them into a corresponding question vector.
6. The yes / no discrimination system based on a clause library according to claim 5, characterized in that, Methods for assigning values to the obtained set of question keywords include: Match the set of question keywords with the keywords in the thesaurus to obtain the corresponding values. Mark the question keywords that do not match as similar keywords. Match the similar keywords with the corresponding similar values and integrate them into the keyword set.
7. The yes / no discrimination system based on a clause library according to claim 6, characterized in that, The method for assigning similarity values based on similar keywords is as follows: Similar keywords are input into a thesaurus for semantic similarity calculation. The highest similarity score is obtained, along with the corresponding topic word's value, which is labeled XS and FZ respectively. The formula is then used to calculate the similarity. Calculate the corresponding similarity assignment, where b3 is the scaling factor, with a value ranging from 0 to 1. <b3≤1, This is the similarity adjustment factor.