Radio and television operation and maintenance platform based on multi-source data fusion

Abstract

The invention discloses a radio and television operation and maintenance platform based on multi-source data fusion. The platform comprises a user question and question classification module; the knowledge base retrieval and query generation module retrieves related contents from the constructed knowledge base according to the classification information; the database query and result summarization module is used for retrieving multi-dimensional data related to the question from the database through the automatically generated query statement; the result analysis and generative text output module is used for converting an analysis result into a natural language through a generative large model on the basis of data summarization and analysis; and the user feedback and intelligent question and answer return module is used for returning the generated answer to the user, the user can continue to ask questions according to the returned result, and the system optimizes and expands the knowledge base and the answer ability through continuous knowledge base updating and feedback learning mechanism. According to the broadcasting and television operation and maintenance platform based on multi-source data fusion, the problem of limitation of application of a general large model in the broadcasting and television field in the prior art is effectively solved.

Description

Technical Field

[0001] The present invention specifically relates to a radio and television operation and maintenance platform based on multi-source data fusion. Background Art

[0002] With the increasing complexity and intelligence requirements of radio and television systems, traditional operation and maintenance methods can no longer meet the needs of efficient and accurate problem diagnosis and decision support. Existing technologies mainly rely on massive text information on the Internet to train general large models, such as the domestic Tongyi Qianwen, Wenxin Yiyan, and foreign ChatGPT. Although these general large models can answer some general technical questions, due to the high broadcast security requirements of radio and television systems, their internal data information and networks are physically isolated, resulting in the existing large model training samples rarely involving professional knowledge in the field of radio and television. Therefore, the existing large models cannot accurately answer technical problems in the field of radio and television, and it is difficult to deal with technical problems such as operation status monitoring, decision support, and intelligent operation and maintenance at the radio and television system level. In order to solve the above problems, there is an urgent need for a question-answering processing platform that can integrate multi-source data for radio and television business systems to improve the automation and intelligence level of radio and television systems in intelligent operation and maintenance, fault diagnosis, and status monitoring. Summary of the invention

[0003] The present invention provides a radio and television operation and maintenance platform based on multi-source data fusion to solve the above-mentioned technical problems, and specifically adopts the following technical solutions:

[0004] A radio and television operation and maintenance platform based on multi-source data fusion, including:

[0005] User question and question classification module, which is used to receive user questions and classify them using semantic understanding technology, dividing them into general questions and radio and television system questions;

[0006] A knowledge base retrieval and query generation module is used to retrieve relevant content from the constructed knowledge base based on classification information. The knowledge base is composed of a knowledge classification table, a knowledge item table, and a knowledge document table. The knowledge base retrieval and query generation module automatically generates query statements to extract information from different data sources;

[0007] The database query and result aggregation module is used to retrieve multi-dimensional data related to the problem from the database through automatically generated query statements, and further process and aggregate the data using long-distance time series analysis and multi-head attention mechanism;

[0008] The result analysis and generative text output module is used to convert the analysis results into natural language through a generative large model based on data aggregation and analysis, generate answers that are easy for users to understand, and provide detailed background information and anomaly analysis;

[0009] The user feedback and intelligent question and answer return module is used to return the generated answers to the user. The user can continue to ask questions based on the returned results. The system optimizes and expands the knowledge base and answering capabilities through continuous knowledge base updates and feedback learning mechanisms.

[0010] Furthermore, the user question and problem classification module improves classification accuracy by combining domain vocabulary and professional terms through sentence modal analysis, divides problems into general problems and broadcasting and television system problems, and classifies broadcasting and television system problems into different categories.

[0011] Furthermore, when processing general problems, the knowledge base retrieval and query generation module directly calls the general large model for processing; when processing radio and television system problems, it combines the radio and television field knowledge base for processing.

[0012] Furthermore, the knowledge base retrieval and query generation module combines structured data and unstructured data, and constructs and optimizes the knowledge base through multi-source data fusion technology to support multi-dimensional analysis and accurate prediction of complex problems.

[0013] Furthermore, the database query and result summary module uses a multi-head attention mechanism for long-distance time series analysis to effectively capture long-term dependencies and important features between data, thereby optimizing the data analysis process.

[0014] Furthermore, the generative large model of the result analysis and generative text output module can combine the processed database information to generate a natural language description that conforms to the context, and convert complex operation and maintenance data into user-friendly text output.

[0015] Furthermore, the user feedback and intelligent question-answering return module continuously expands and optimizes the depth and breadth of the knowledge base through user feedback and system learning, and utilizes the feedback learning mechanism to enhance the model's adaptive ability and problem-solving ability.

[0016] Furthermore, the radio and television operation and maintenance platform based on multi-source data fusion integrates various types of information such as operation data, monitoring indicators, operation and maintenance records and related professional documents of the radio and television system into a knowledge base through multi-source data fusion technology, and provides real-time monitoring, dynamic warning and intelligent analysis functions.

[0017] Furthermore, the radio and television operation and maintenance platform based on multi-source data fusion adopts field-specific data to pre-train and fine-tune large models in response to special needs in the radio and television field, ensuring that the model can understand and handle technical problems unique to the radio and television system.

[0018] Furthermore, the radio and television operation and maintenance platform based on multi-source data fusion adopts semantic understanding technology to classify user questions, and combines pre-training and fine-tuning of large language models to improve the accuracy and intelligence level of question classification.

[0019] The present invention provides a radio and television operation and maintenance platform based on multi-source data fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 It is a schematic diagram of a radio and television operation and maintenance platform based on multi-source data fusion of the present invention. DETAILED DESCRIPTION

[0022] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0023] The radio and television question and answer processing platform with multi-source data fusion characteristics of the present invention builds an efficient and intelligent operation and maintenance platform by integrating technologies such as multi-source data fusion, generative large models, and knowledge base construction.

[0024] like Figure 1 The present application shows a radio and television operation and maintenance platform with multi-source data fusion features, including: user question and problem classification module, knowledge base retrieval and query generation module, database query and result summary module, result analysis and generative text output module and user feedback and intelligent question and answer return module.

[0025] Specifically, the user question and question classification module is used to receive user questions and classify questions using semantic understanding technology, dividing questions into general questions and radio and television system questions. The knowledge base retrieval and query generation module is used to retrieve relevant content from the constructed knowledge base based on classification information. The knowledge base is composed of a knowledge classification table, a knowledge entry table, and a knowledge document table. The knowledge base retrieval and query generation module automatically generates query statements to extract information from different data sources. The database query and result summary module is used to retrieve multi-dimensional data related to the question from the database through automatically generated query statements, and further process and summarize the data using long-distance time series analysis and multi-head attention mechanisms. The result analysis and generative text output module is used to convert the analysis results into natural language through a generative large model based on data aggregation and analysis, generate answers that are easy for users to understand, and provide detailed background information and abnormal analysis. The user feedback and intelligent question and answer return module is used to return the generated answers to the user. The user can continue to ask questions based on the returned results. The system optimizes and expands the knowledge base and answer capabilities through continuous knowledge base updates and feedback learning mechanisms.

[0026] In the implementation mode of the present application, the radio and television operation and maintenance platform with multi-source data fusion features integrates various types of information such as operation data, monitoring indicators, operation and maintenance records and related professional documents of the radio and television system into a knowledge base through multi-source data fusion technology, providing real-time monitoring, dynamic warning and intelligent analysis functions.

[0027] In the implementation mode of the present application, the radio and television operation and maintenance platform with multi-source data fusion features uses field-specific data to pre-train and fine-tune large models to meet the special needs of the radio and television field, ensuring that the model can understand and handle technical problems unique to the radio and television system.

[0028] The functions of the above modules are further described below.

[0029] User question and problem classification module: Users submit questions to the system through the platform. The questions may involve various status monitoring, fault diagnosis, operation records, etc. of the radio and television system. After receiving the user's question, the platform uses semantic understanding technology to classify the questions into general questions and radio and television system questions. Specifically, the user question and problem classification module improves the accuracy of classification through sentence modal analysis combined with domain vocabulary and professional terms, and classifies the questions into general questions and radio and television system questions, and classifies radio and television system questions into different categories.

[0030] For general questions, the system directly calls the general large model for processing and generates answers. For radio and television system issues, the system combines the radio and television field knowledge base for processing and generates answers. In the implementation mode of the present application, the radio and television operation and maintenance platform with multi-source data fusion features uses semantic understanding technology to classify user questions, combined with the pre-training and fine-tuning of the large language model to improve the accuracy and intelligence level of question classification. This classification process is optimized through the pre-training and fine-tuning of the large language model to ensure that the intent and context of the question can be accurately understood. For example, the question "Which sites in the system have the most abnormalities in the past year?" will be classified into the "System Status Analysis" category, and the system will retrieve relevant content from the operation and maintenance knowledge base based on this classification. By fine-tuning the large language model, it can be adapted to the operation and maintenance problems unique to the radio and television field. Sentence modal analysis combines domain vocabulary and professional terminology to improve the accuracy of classification.

[0031] Knowledge base retrieval and query generation module: In the implementation of the present application, the knowledge base retrieval and query generation module directly calls the general large model for processing when processing general problems. When processing radio and television system problems, the radio and television field knowledge base is combined for processing. Once the problem is correctly classified, the system will retrieve relevant content from the constructed knowledge base based on the classification information. The knowledge base consists of three core data tables: knowledge classification table, knowledge entry table and knowledge document table. The knowledge classification table helps the system to quickly locate the scope of the problem, the knowledge entry table provides detailed information about the problem, and the knowledge document table associates related documents to help the system provide in-depth support for complex problems. In order to efficiently handle user questions, the system will automatically generate query statements. These query statements extract information from different data sources according to the category of user questions. For example, for "system status analysis" type questions, the query statement will extract system monitoring indicators, alarm records and historical operation data. Combining structured data (such as monitoring indicators) and unstructured data (such as technical documents, operation manuals, etc.), the knowledge base is constructed and optimized through multi-source data fusion technology to support multi-dimensional analysis and accurate prediction of complex problems.

[0032] In the implementation manner of the present application, the knowledge base retrieval and query generation module combines structured data and unstructured data, and constructs and optimizes the knowledge base through multi-source data fusion technology to support multi-dimensional analysis and accurate prediction of complex problems.

[0033] Database query and result summary module: Through automatically generated query statements, the system retrieves multi-dimensional data related to the problem from the database. These data include system status, fault records, alarm data, etc., and the data is further processed and summarized using long-distance time series analysis and multi-head attention mechanisms. The multi-head attention mechanism of long-distance time series analysis is used to effectively capture the long-term dependencies and important features between data and optimize the data analysis process. This enables the system to comprehensively consider the historical and real-time data of the system status to generate accurate operation and maintenance analysis reports. For example, for a problem involving system status monitoring, the system will analyze the monitoring data within the past year, combine the real-time data, identify key abnormal patterns through the multi-head attention mechanism, and generate a detailed analysis report.

[0034] In an implementation of the present application, the database query and result aggregation module uses a multi-head attention mechanism for long-distance time series analysis to effectively capture long-term dependencies and important features between data, thereby optimizing the data analysis process.

[0035] Result analysis and generative text output module: Based on data aggregation and analysis, the system converts the analysis results into natural language through a generative big model to generate answers that are easy for users to understand. The system not only provides numerical results, but also combines historical data, system status and other information to generate detailed background descriptions and exception analysis. For example, when answering "Which sites in the system have the most exceptions in the past year?", the system will list the sites with the most exceptions, and describe the types of exceptions, frequency of occurrence and possible causes. Generative big models play an important role here to ensure that the answers are not only accurate, but also smooth and natural. Generative big models can combine processed database information to generate natural language descriptions that conform to the context. This technology enables the system to convert complex operation and maintenance data into user-friendly text output to enhance the user experience.

[0036] In the implementation manner of the present application, the generative large model of the result analysis and generative text output module can combine the processed database information to generate a natural language description that conforms to the context, and convert complex operation and maintenance data into user-friendly text output.

[0037] User feedback and intelligent question-answering return module: The system returns the generated answers to the user, and the user can continue to ask questions based on the returned results. The system can optimize and expand its knowledge base and answering capabilities through continuous knowledge base updates and feedback learning mechanisms. Every time a user asks a question, the system will not only return an immediate answer, but also update the system based on user feedback to further improve the quality of question-answering. User feedback information is integrated into the knowledge base and becomes an important basis for answering future questions. The knowledge base is iteratively updated, and the depth and breadth of the knowledge base are continuously expanded and optimized through user feedback and system learning. The feedback learning mechanism is used to improve the model's adaptive ability and problem-solving ability.

[0038] In the implementation mode of the present application, the user feedback and intelligent question and answer return module continuously expands and optimizes the depth and breadth of the knowledge base through user feedback and system learning, and utilizes the feedback learning mechanism to enhance the model's adaptive ability and problem-solving ability.

[0039] Through the above implementation methods, the radio and television question and answer processing platform with multi-source data fusion characteristics of the present invention can effectively solve the limitations of the application of general large models in the radio and television field in the prior art, provide all-round technical support for the intelligent operation and maintenance of the radio and television system, improve the accuracy and efficiency of intelligent operation and maintenance, optimize problem classification and response capabilities, realize intelligent data analysis and generative question and answer, and enhance user experience.

[0040] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A radio and television operation and maintenance platform based on multi-source data fusion, characterized in that: include: User question and question classification module, which is used to receive user questions and classify them using semantic understanding technology, dividing them into general questions and radio and television system questions; A knowledge base retrieval and query generation module is used to retrieve relevant content from the constructed knowledge base based on classification information. The knowledge base is composed of a knowledge classification table, a knowledge item table, and a knowledge document table. The knowledge base retrieval and query generation module automatically generates query statements to extract information from different data sources; The database query and result aggregation module is used to retrieve multi-dimensional data related to the problem from the database through automatically generated query statements, and further process and aggregate the data using long-distance time series analysis and multi-head attention mechanism; The result analysis and generative text output module is used to convert the analysis results into natural language through a generative large model based on data aggregation and analysis, generate answers that are easy for users to understand, and provide detailed background information and anomaly analysis; The user feedback and intelligent question and answer return module is used to return the generated answers to the user. The user can continue to ask questions based on the returned results. The system optimizes and expands the knowledge base and answering capabilities through continuous knowledge base updates and feedback learning mechanisms.

2. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1 is characterized in that: The user question and question classification module improves the accuracy of classification by combining domain vocabulary and professional terms through sentence modal analysis, divides questions into general questions and radio and television system questions, and classifies radio and television system questions into different categories.

3. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1 is characterized in that: When processing general problems, the knowledge base retrieval and query generation module directly calls the general large model for processing; when processing radio and television system problems, it combines the radio and television field knowledge base for processing.

4. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1 is characterized in that: The knowledge base retrieval and query generation module combines structured data and unstructured data, and constructs and optimizes the knowledge base through multi-source data fusion technology to support multi-dimensional analysis and accurate prediction of complex problems.

5. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1 is characterized in that: The database query and result summary module uses a multi-head attention mechanism for long-distance time series analysis to effectively capture long-term dependencies and important features between data and optimize the data analysis process.

6. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1, characterized in that: The generative large model of the result analysis and generative text output module can combine the processed database information to generate a natural language description that conforms to the context, and convert complex operation and maintenance data into user-friendly text output.

7. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1 is characterized in that: The user feedback and intelligent question-answering return module continuously expands and optimizes the depth and breadth of the knowledge base through user feedback and system learning, and utilizes the feedback learning mechanism to enhance the model's adaptive ability and problem-solving ability.

8. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1, characterized in that: The radio and television operation and maintenance platform based on multi-source data fusion integrates various types of information such as operation data, monitoring indicators, operation and maintenance records and related professional documents of the radio and television system into a knowledge base through multi-source data fusion technology, and provides real-time monitoring, dynamic warning and intelligent analysis functions.

9. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1, characterized in that: The radio and television operation and maintenance platform based on multi-source data fusion adopts field-specific data to pre-train and fine-tune large models in response to special needs in the radio and television field, ensuring that the model can understand and handle technical problems unique to the radio and television system.

10. The radio and television operation and maintenance platform based on multi-source data fusion according to claim 1, characterized in that: The radio and television operation and maintenance platform based on multi-source data fusion adopts semantic understanding technology to classify user questions, and combines pre-training and fine-tuning of a large language model to improve the accuracy and intelligence level of question classification.

Images