Highway maintenance knowledge question and answer method and system based on multi-modal information

Abstract

The application relates to the technical field of highway intelligent maintenance, and discloses a highway maintenance knowledge question and answer method and system based on multi-modal information, which comprises a multi-dimensional acquisition module and an intelligent management module. The system obtains all highway maintenance knowledge and all user question data through the multi-dimensional acquisition module, and classifies and forms a data set, the intelligent management module analyzes entities and connection relationships in the highway maintenance knowledge, constructs a knowledge graph, the multi-dimensional integration has high accuracy, the intelligent management module analyzes different forms of question content, queries corresponding target triples in the knowledge graph, preferentially returns the result with the strongest correlation, and counts the query duration, analyzes the reply speed and quality score of each user when asking questions, evaluates the query efficiency and query quality of the highway maintenance knowledge, and outputs corresponding query instructions, so that the intelligent and professional answer of intelligent maintenance is realized, and the knowledge query efficiency is high.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent highway maintenance technology, in particular to a highway maintenance knowledge question and answer method and system based on multi-modal information. BACKGROUND

[0002] With the continuous growth of highway mileage, the importance of highway maintenance work is increasingly prominent. In order to cope with the complex road network structure and diversified maintenance needs, the industry technical standards and knowledge system are constantly iterated and upgraded. Various maintenance technical standards, maintenance implementation suggestions and other contents are increasing, and data is scattered in text reports, sensors, images and other different modalities, and information integration efficiency is low. Traditional question and answer systems cannot understand professional terms and context, resulting in insufficient response accuracy, lagging maintenance knowledge update, lack of dynamic learning mechanism, and difficulty in timely adapting to new highway maintenance standards.

[0003] In solving these problems, there are many difficulties, one of which is the bottleneck of professional term understanding. Highway maintenance involves professional terms such as crack types, material ratios, and process standards. The accuracy of general NLP models may be less than 60%. The second is the difficulty of adapting to multi-modal scenarios. Maintenance personnel often need to describe diseases through pictures or mark repair areas on CAD drawings. Existing single-text question and answer models cannot handle it. The third is the challenge of non-structured data dispersion and regional standard differences in the process of data acquisition and knowledge base construction. The fourth is the lag in standard updates. The update frequency of maintenance standards is 3-5% per year, while the update cycle of traditional knowledge bases needs to be compressed from 3 months to 7 days. This situation makes it an important problem to develop an AI intelligent question and answer robot that can cover more than 80% of industry high-frequency questions, support dynamic knowledge update, support multi-round follow-up questions and context understanding, and support voice / picture input. SUMMARY

[0004] In view of the shortcomings of the prior art, the present application provides a highway maintenance knowledge question and answer method and system based on multi-modal information, which has the advantages of high multi-dimensional integration accuracy and high knowledge query efficiency, and solves the problems of low information integration efficiency, slow update lag response speed of traditional highway maintenance knowledge question and answer systems.

[0005] To achieve the above purpose, the present application provides the following technical scheme: a highway maintenance knowledge question and answer method based on multi-modal information, comprising the following steps:

[0006] Step 1: Connect the database and big data platform through the network, obtain all highway maintenance knowledge and all user question data, and classify them into knowledge data set and question data set;

[0007] Step two: according to the professional data set, analyze the entity and connection relationship in the highway maintenance knowledge, and construct the corresponding knowledge graph ;

[0008] Step three: according to the question data set, analyze the different forms of question content, query the corresponding target triple in the knowledge graph , and count the corresponding query time ; ;

[0009] Step four: according to the question data set and query time , analyze the reply speed and quality score of each user when answering questions ; ;

[0010] Step five: set a fixed value of speed threshold , combined with the reply speed and quality score , evaluate the query efficiency and query quality of highway maintenance knowledge, and output the corresponding query instruction.

[0011] Preferably, in step one, the knowledge data set includes maintenance specifications, technical standards, operation manuals, historical maintenance reports, test reports, expert experience documents and research papers.

[0012] Preferably, in step two, the question data set includes the text content, voice content and image content of each user.

[0013] Preferably, in step two, the knowledge graph construction process is as follows:

[0014] S11, through natural language processing technology, identify the entity in the knowledge data set, the entity includes roadbed, pavement, bridge, tunnel, safety facilities, green plants and emergency supplies;

[0015] S12, count the number of entities in the knowledge data set, and then set the corresponding node according to each entity, and the number of nodes is consistent with the number of entities;

[0016] S13, through natural language processing technology, analyze the connection relationship between each entity in the knowledge data set, and then set the corresponding edge according to the connection relationship between each entity;

[0017] S14, integrate triplets through natural language processing technology, arrange all entities and connection relationships in S11-S13 in the form of "entity→relationship→entity", and save them in the form of "node→edge→node";

[0018] If a single entity has a connection relationship with multiple entities, it is arranged in the form of "entity→relationship→entity→relationship→entity", and saved in the form of "node→edge→node→edge→node";

[0019] knowledge graph is composed of a plurality of triples.

[0020] Preferably, in step three, the target triple The query process is as follows:

[0021] S21, according to the question data set, extract the first user's question data, and mark the first user's question content each time as , to represent the first to the first question content of the user;

[0022] S22, according to the input form of the first user's question content, query the corresponding target triple , as follows:

[0023] If the first question content of the first user is in text form, the entity in the question content should be identified by using natural language processing technology;

[0024] If the first question content of the first user is in voice form, the entity in the input question content should be identified by using voice processing software;

[0025] If the first question content of the first user is in image form, the entity in the input question content should be identified by using image processing software;

[0026] Mark the entity in the first question content of the first user to the entity in the first question content of the first user as ;

[0027] S23, in the knowledge graph , query all triples containing entity , arrange them in order from many to few according to the number of entities containing, and mark them as , to In knowledge graph In the middle, the first group to the... Group contains entities The triplet, where the first triplet Contains entities The largest number, the first group of triplets That is, the target triplet .

[0028] Preferably, in step four, the recovery speed The calculation process is as follows:

[0029]

[0030] In the formula, Indicates the first Query duration per user Indicates the first Number of questions asked by each user Indicates reply to the first The time required for each user to answer each question is the [number]th [question]. Response speed when users ask questions .

[0031] Preferably, in step four, the quality score is... The evaluation process is as follows:

[0032] If the first The first question asked by a user Entities in To the The content of the question Entities in If all are different, it means that the target triplet of the output is different. It accurately meets users' needs for information on highway maintenance. The first [item / section] should be [presumably the correct one / section]. Quality rating of user Q&A It is recorded as 1 point;

[0033] If the first The first question asked by a user Entities in To the The content of the question Entities in If all are identical, it means the target triplet to be output is... This does not accurately meet users' needs for information on highway maintenance; the first [item] should be [addressed / reproduced]. Quality rating of user Q&A It is recorded as 0 points.

[0034] Preferably, the step five, the reply speed ≥ speed threshold , indicating that the query efficiency is low, and the user should be suggested to simplify the input content to reduce the query range in the knowledge graph .

[0035] Preferably, the step five, the quality score is 0, indicating that the query quality is low, and the natural language processing technology should be used to mine the near-synonymous entities to expand the query range in the knowledge graph .

[0036] The highway maintenance knowledge question and answer system based on multi-modal information comprises a multi-dimensional acquisition module and an intelligent management module.

[0037] The multi-dimensional acquisition module is composed of a professional data unit and a user data unit, the professional data unit acquires a knowledge data set by connecting a database through a network, the knowledge data set comprises all highway maintenance knowledge, and the user data unit acquires a question data set by connecting a big data platform through a network, the question data set comprises all user question data.

[0038] The intelligent management module is composed of a graph analysis unit, a knowledge query unit, a reply evaluation unit and a question and answer management unit, the graph analysis unit analyzes entities and connection relationships in highway maintenance knowledge according to the professional data set, and constructs a corresponding knowledge graph , the knowledge query unit analyzes different forms of question content according to the question data set, queries corresponding target triples in the knowledge graph , and counts corresponding query time lengths , the reply evaluation unit analyzes the reply speed and the quality score of each user during question and answer according to the question data set and the query time length , the question and answer management unit is provided with a fixed numerical speed threshold , in combination with the reply speed and the quality score , evaluates the query efficiency and the query quality of the highway maintenance knowledge, and outputs corresponding query instructions.

[0039] Compared with the prior art, the present application provides a highway maintenance knowledge question and answer method and system based on multi-modal information, which has the following beneficial effects:

[0040] ​1. The application connects the database and the big data platform through the multi-dimensional acquisition module network, obtains all highway maintenance knowledge and all user question data, and classifies and forms knowledge data sets and question data sets, and the intelligent management module analyzes entities and connection relationships in the highway maintenance knowledge according to professional data sets, and constructs a corresponding knowledge graph , converts unstructured text into a semantic association network, upgrades the query from ''keyword matching'' to ''semantic association search'', greatly improves the accuracy, effectively solves the problems of scattered traditional maintenance knowledge and difficult retrieval, avoids the operation of repeated query of users among different systems, and significantly reduces the data acquisition time, and the multi-dimensional integration has high accuracy.

[0041] 2. The application analyzes different forms of question content in the knowledge graph according to the question data set through the intelligent management module, queries the corresponding target triplets , preferentially returns the result with the strongest association, and statistically analyzes the corresponding query duration , effectively adapts to the disease identification needs of field workers, reduces the use threshold, and analyzes the reply speed and quality score of each user when asking and answering according to the question data set and the query duration , sets a fixed numerical speed threshold , and combines the reply speed and quality score to evaluate the query efficiency and query quality of the highway maintenance knowledge, and output corresponding query instructions, realize the intelligent and professional answer of intelligent maintenance, and the knowledge query efficiency is high. BRIEF DESCRIPTION OF DRAWINGS

[0042] Fig. 1 is a method step diagram of the application;

[0043] Fig. 2 is a system flowchart of the application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0045] Because the traditional highway maintenance knowledge question and answer system has low information integration efficiency, slow update lag response speed, therefore, please refer to Figs. 1-2The application provides a highway maintenance knowledge question and answer method and system based on multi-modal information, and specifically as follows:

[0046] The highway maintenance knowledge question and answer method based on multi-modal information comprises the following steps:

[0047] Step one: connect the database and the big data platform through the network, obtain all highway maintenance knowledge and all user question data, and classify and form a knowledge data set and a question data set;

[0048] The knowledge data set comprises maintenance specifications, technical standards, operation manuals, historical maintenance reports, detection reports, expert experience documents and research papers;

[0049] Specifically, the data attributes of the highway maintenance specifications, technical standards and operation manuals are generally structured data, and the data attributes of the highway maintenance historical maintenance reports, detection reports, expert experience documents and research papers are generally unstructured data, the structured data and the unstructured data are fused and analyzed by using the natural language processing technology, and the compatibility of the knowledge graph is ensured;

[0050] The question data set comprises the text content, the voice content and the image content of each user;

[0051] Step two: analyze the entities and the connection relationships in the highway maintenance knowledge according to the professional data set, and construct a corresponding knowledge graph ;

[0052] The construction process of the knowledge graph is as follows:

[0053] S11, the entities in the knowledge data set are recognized by using the natural language processing technology, and the entities comprise roadbeds, pavements, bridges, tunnels, safety facilities, green plants and emergency supplies;

[0054] S12, the number of the entities in the knowledge data set is counted, and then the corresponding nodes are set according to each entity, and the number of the nodes is consistent with the number of the entities;

[0055] S13, the connection relationships between each entity in the knowledge data set are analyzed by using the natural language processing technology, and then the corresponding edges are set according to the connection relationships between each entity;

[0056] S14, all the entities and the connection relationships in S11 to S13 are arranged in the form of “entity→relationship→entity” and saved in the form of “node→edge→node” by integrating the triples by using the natural language processing technology;

[0057] If a single entity has connections with multiple entities, they are arranged in the form of "entity → relation → entity → relation → entity" and then saved in the form of "node → edge → node → edge → node".

[0058] knowledge graph It consists of several triples;

[0059] Specifically, knowledge graph Transforming unstructured text into a semantic association network upgrades queries from "keyword matching" to "semantic association search," significantly improving accuracy. This effectively solves the problems of scattered traditional maintenance knowledge and difficulty in retrieval, avoids users repeatedly querying across different systems, and significantly reduces data acquisition time. The multi-dimensional integration has high accuracy.

[0060] Step 3: Based on the question dataset, analyze the content of different question formats and apply them to the knowledge graph. Query the corresponding target triplet in the middle And count the corresponding query duration. ;

[0061] Target Triples The query process is as follows:

[0062] S21. Based on the question dataset, extract the first... The question data of the first user, and the first user's question data .... Each user's question is marked as , to Indicates the first The first time a user reached the The content of the next question;

[0063] S22, according to the first The query retrieves the target triplet based on the input format of a user's question. The details are as follows:

[0064] If the first The first question asked by a user In text form, natural language processing technology should be used to identify entities in the question content;

[0065] If the first The first question asked by a user If the input is in voice format, voice processing software should be used to recognize the entities in the input question.

[0066] If the first The first question asked by a user If the input is in image format, image processing software should be used to identify the entities in the input question.

[0067] The entity in the first question content of the first user is marked as the entity in the second question content of the first user. The entity in the first question content of the first user is marked as the entity in the second question content of the first user. The entity in the first question content of the first user is marked as the entity in the second question content of the first user. ;

[0068] Specifically, the corresponding target triplets are queried according to the input form of the user's question , which realizes effective adaptation to the disease identification needs of the on-site staff and reduces the use threshold;

[0069] S23、In the knowledge graph , query all triplets containing entity , arrange them from many to few according to the number of contained entities , and mark them as , to indicate that in the knowledge graph , the first group to the group of triplets containing entity , wherein the first group of triplets contains the most entities , and the first group of triplets is the target triplet ;

[0070] Specifically, the target triplets are sorted according to the frequency of entity occurrence , and the result with the strongest relevance is returned preferentially;

[0071] Step four: according to the question data set and the query duration , analyze the reply speed and quality score of each user when asking questions;

[0072] The reply speed calculation process is as follows:

[0073]

[0074] In the formula, represents the query duration of the first user, represents the number of questions of the first user, represents the time required to reply to each question of the first user, that is, the reply speed of the first user when asking questions, which objectively reflects the system efficiency; ​​

[0075] quality score The evaluation process is as follows:

[0076] If the entities in the first question content of the first user are different from the entities in the question content from the second to the nth, it indicates that the output target triple accurately meets the user's query demand in the expressway maintenance knowledge, and the quality score of the first user in the question and answer process should be recorded as 1 point.

[0077] If the entities in the first question content of the first user are different from the entities in the question content from the second to the nth, it indicates that the output target triple accurately meets the user's query demand in the expressway maintenance knowledge, and the quality score of the first user in the question and answer process should be recorded as 1 point.

[0078] Step five: set a fixed numerical speed threshold , combined with the reply speed and the quality score , evaluate the query efficiency and query quality of the expressway maintenance knowledge, and output the corresponding query instruction.

[0079] When the reply speed is greater than or equal to the speed threshold , it indicates that the query efficiency is low, and the user should be advised to simplify the input content and focus on the core entity to narrow the query range in the knowledge graph .

[0080] When the quality score is 0, it indicates that the query quality is low, and the near-synonymous entity should be mined using natural language processing technology to expand the query range in the knowledge graph .

[0081] Specifically, the content of the corresponding maintenance knowledge and daily maintenance data is integrated and analyzed to answer, realizing the intelligent and professional answer of intelligent maintenance, improving the maintenance knowledge query efficiency, query quality and knowledge query timeliness, and improving the user experience and work efficiency.

[0082] ​​​​​​​​​​​​​​​​​​A highway maintenance knowledge Q&A system based on multimodal information, including a multidimensional data acquisition module and an intelligent management module;

[0083] The multidimensional data acquisition module consists of a professional data unit and a user data unit. The professional data unit collects knowledge datasets by connecting to a database via a network. The knowledge datasets include all highway maintenance knowledge. The user data unit collects question datasets by connecting to a big data platform via a network. The question datasets include all user question data.

[0084] The intelligent management module consists of a graph analysis unit, a knowledge query unit, a response evaluation unit, and a question-and-answer management unit. The graph analysis unit analyzes entities and connections in highway maintenance knowledge based on professional datasets and constructs a corresponding knowledge graph. The knowledge query unit analyzes different forms of question content based on the question dataset and applies it to the knowledge graph. Query the corresponding target triplet in the middle And count the corresponding query duration. The multidimensional integration achieves high accuracy, and the response evaluation unit is based on the question dataset and query duration. Analyze the response speed of each user when asking and answering questions. and quality rating The question-and-answer management unit has a fixed speed threshold. Combined with response speed and quality rating It evaluates the efficiency and quality of knowledge retrieval for highway maintenance and outputs corresponding query instructions, demonstrating high knowledge retrieval efficiency.

[0085] Example 1

[0086] In this experiment, a user who asked 10 questions within one hour was selected as the experimental subject. Statistics were compiled on the query time of this user on the highway maintenance knowledge Q&A system. The response time is 110 seconds. The calculation process is as follows:

[0087]

[0088] In the formula, This indicates the time required to respond to each of the user's questions, i.e., the response speed when answering questions from that user. speed threshold The response time was set to 10 seconds, which, based on the assessment, determined the response speed of the highway maintenance knowledge Q&A system when the user asked the question. ≥ speed threshold This indicates low query efficiency; the user should be advised to simplify their input to narrow down the knowledge graph. query range in the query.

[0089] Embodiment 2

[0090] In this experiment, the Highway Asphalt Pavement Maintenance Technical Specification is selected as the experimental object, and the knowledge graph The construction process is as follows:

[0091] S11, through natural language processing technology, the entities in the Highway Asphalt Pavement Maintenance Technical Specification are identified, including roadbed, pavement, bridge, tunnel, safety facilities and emergency supplies;

[0092] S12, the number of entities in the Highway Asphalt Pavement Maintenance Technical Specification is counted, and then the corresponding nodes are set according to each entity, and the number of nodes is consistent with the number of entities;

[0093] The nodes include roadbed, pavement, bridge, tunnel, safety facilities and emergency supplies.

[0094] S13, through natural language processing technology, the connection relationship between each entity in the Highway Asphalt Pavement Maintenance Technical Specification, and then according to the connection relationship between each entity, the corresponding edge is set;

[0095] The edges include the connection relationship between the roadbed and the pavement, the connection relationship between the bridge and the pavement, the connection relationship between the tunnel and the pavement, and the connection relationship between the safety facilities and the pavement.

[0096] S14, through natural language processing technology, integrate triples, arrange all entities and connection relationships in S11 to S13 in the form of "entity→relationship→entity", and save them in the form of "node→edge→node";

[0097] If a single entity has a connection relationship with multiple entities, arrange them in the form of "entity→relationship→entity→relationship→entity", and save them in the form of "node→edge→node→edge→node";

[0098] The triples include roadbed→located at bottom→pavement, bridge→located at bottom→pavement, tunnel→connected to each other→pavement, and safety facilities→located on both sides→pavement.

[0099] Knowledge graph Composed of several triples.

[0100] The size of the threshold is set for easy comparison. The size of the threshold depends on the amount of sample data and the base number set by the person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameters and the quantized values.

[0101] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value, and the coefficients in the formula are set by the person skilled in the art according to the actual situation. The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for Freeway Maintenance Knowledge Question Answering based on multi-modal information, characterized in that, Comprise the following steps: Step one: through the network connection database and big data platform, get all the highway maintenance knowledge and all user question data, and classified into knowledge data set and question data set; Step two: According to the professional dataset, analyze the entities and connection relationships in the highway maintenance knowledge, and build the corresponding knowledge graph ; Knowledge graph The construction process is as follows: S11, through natural language processing technology, identify the entity in the knowledge data set, the entity includes roadbed, pavement, bridge, tunnel, safety facilities, green plants and emergency supplies; S12, statistics entity quantity in knowledge data set, then according to each entity, set the corresponding node, and the number of nodes and the number of entities are consistent; S13, through natural language processing technology, analyze the connection between each entity in the knowledge data set, and set the corresponding edge according to the connection between each entity; S14, through natural language processing technology integration triple, all entities and connection in S11 to S13 are arranged in the form of "entity→relation→entity", and saved in the form of "node→edge→node"; If a single entity has a connection relationship with multiple entities, it is arranged in the form of "entity→relation→entity→relation→entity", and saved in the form of "node→edge→node→edge→node"; Knowledge graph consists of several triples; Step three: According to the question dataset, analyze different forms of question content, query the corresponding target triple in the knowledge graph , and count the corresponding query time . ​ Target triplets The query flow is as follows: S21. Based on the question dataset, extract the first... The question data of the first user, and the first user's question data .... Each user's question is marked as , to Indicates the first The first time a user reached the The content of the next question; S22, according to the first The query retrieves the target triplet based on the input format of a user's question. The details are as follows: If the first The first question asked by a user In text form, natural language processing technology should be used to identify entities in the question content; If the first question content of the user is in voice form, voice processing software should be used to recognize the entity in the input question content. ​​ If the first question content of the user is in the form of an image, image processing software should be used to identify the entities in the input question content. If the first question content of the user is in the form of an image, image processing software should be used to identify the entities in the input question content.​ mark the entity in the first time question content of the first user to the entity in the second question content ; and mark the entity in the first S23、In the knowledge graph , query all triples containing the entity , arrange them from more to less according to the number of containing entities , and mark them as , to represent that in the knowledge graph , the first group to the group of triples containing the entity , wherein the first group of triples contains the most entities , and the first group of triples is the target triple ; Step four: Analyzing the response speed and quality score of each user when asking and answering questions according to the question dataset and query duration ;​​ Recovery speed The calculation proceeds as follows: In the formula, represents the query duration of the th user, represents the number of questions of the th user, represents the time required to reply to each question of the th user, that is, the reply speed when the th user asks and answers questions ; Quality score The evaluation procedure was as follows: If the first The first question asked by a user Entities in To the The content of the next question Entities in If all are different, it means that the target triplet of the output is different. It accurately meets users' needs for information on highway maintenance. The first [item / section] should be [presumably the correct one / section]. Quality rating of user Q&A It is recorded as 1 point; If the first The first question asked by a user Entities in To the The content of the question Entities in If all are identical, it means the target triplet to be output is... This does not accurately meet users' needs for information on highway maintenance; the first [item] should be [addressed / reproduced]. Quality rating of user Q&A It is recorded as 0 points; Step five: set a fixed value of speed threshold , combined with the reply speed and the quality score , to evaluate the query efficiency and query quality of highway maintenance knowledge and output the corresponding query instructions.

2. The Freeway Maintenance Knowledge Q&A method based on multi-modal information according to claim 1, characterized in that: In the step one, the knowledge data set includes maintenance specification, technical standard, operation manual, historical maintenance report, detection report, expert experience document and research paper.

3. The Freeway Maintenance Knowledge Q&A method based on multi-modal information according to claim 2, characterized in that: In the step one, the question data set includes the text content, voice content and image content of each user.

4. The Freeway Maintenance Knowledge Q&A method based on multi-modal information according to claim 3, characterized in that: The step five, the reply speed ≥ speed threshold value When, it indicates that the query efficiency is low, and the user should be suggested to simplify the input content to reduce the query range in the knowledge graph .

5. The Freeway Maintenance Knowledge Q&A method based on multi-modal information according to claim 4, characterized in that: The step five, quality score When the score is 0, it indicates that the query quality is low, and natural language processing technology should be used to mine near-synonymous entities to expand the query range of the knowledge graph. ​ 6. The highway maintenance knowledge Q&A system based on multi-modal information, applied to the highway maintenance knowledge Q&A method based on multi-modal information in any one of claims 1-5, characterized in that: Comprise multi-dimensional acquisition module and intelligent management module; The multi-dimensional acquisition module is composed of professional data unit and user data unit, the professional data unit acquires knowledge data set by connecting database through network, the knowledge data set includes all the highway maintenance knowledge, the user data unit acquires question data set by connecting big data platform through network, the question data set includes all user question data; The intelligent management module consists of a graph analysis unit, a knowledge query unit, a response evaluation unit, and a question-and-answer management unit. The graph analysis unit analyzes entities and connections in highway maintenance knowledge based on professional datasets and constructs a corresponding knowledge graph. The knowledge query unit analyzes different forms of question content based on the question dataset and applies it to the knowledge graph. Query the corresponding target triplet in the middle And count the corresponding query duration. The response evaluation unit evaluates the responses based on the query dataset and the query duration. Analyze the response speed of each user when asking and answering questions. and quality rating The question-and-answer management unit is set with a fixed speed threshold value. Combined with response speed and quality rating The system evaluates the efficiency and quality of information retrieval related to highway maintenance and outputs corresponding query commands.

Images