A rural multi-source data fusion analysis display method and system and a medium

Abstract

The application discloses a kind of rural multi-source data fusion analysis display method, system and medium, belong to data protection technical field, for solving the lack of village multi-source heterogeneous data structured fusion technology in prior art, lack of targeted generative question and answer agent, the accuracy of village information query and user satisfaction are low Problem. Including: constructing structured village multi-source data set;Obtain the historical question data of village information query website, and divide historical question data according to query content, obtain multiple question types;Each historical question data is associated with the corresponding answer and is annotated, and multiple types of question and answer sub-bases are constructed according to the question type belonging to, to form a classified retrieval enhanced knowledge base;Construct village question and answer generative model, and train the model by the data in retrieval enhanced knowledge base;The village question and answer generative model after training is user feedback tuning, to obtain the final village multi-source data fusion analysis model.

Description

Technical Field

[0001] This invention relates to the field of data protection technology, and in particular to a method, system, and medium for the fusion analysis and display of multi-source data in rural areas. Background Technology

[0002] With the advancement of the digital village strategy, digital twin technology has gradually become a core technical means for the digital protection and governance of traditional villages. By digitally collecting, modeling and mapping physical villages, it enables dynamic perception and simulation analysis of village status.

[0003] Existing digital twin technologies for villages primarily focus on building preservation and safety assessment scenarios. However, their data coverage is limited, confining them to building physical parameters and safety monitoring data. They fail to structurally integrate multi-source, heterogeneous data such as basic village information, industry-related information, and cultural and geographical information, thus failing to support users' diverse information needs. Furthermore, existing technologies largely serve management-side monitoring, assessment, and protection control, lacking a retrieval-enhanced knowledge base for village information queries, making it difficult to efficiently respond to users' natural language queries. Additionally, general generative models exhibit low accuracy and information matching in village information query scenarios and lack an integrated intelligent display platform. Multi-source village data is stored and displayed independently, preventing users from accessing comprehensive village information through a unified portal, resulting in poor information retrieval efficiency and a suboptimal user experience. Summary of the Invention

[0004] This invention provides a method, system, and medium for the fusion analysis and display of multi-source data in rural areas, which addresses the following technical problems: existing technologies lack structured fusion technology for multi-source heterogeneous data in villages, lack targeted generative question-answering agents, and have low accuracy and user satisfaction in village information queries.

[0005] The embodiments of the present invention adopt the following technical solutions: On the one hand, embodiments of the present invention provide a method for the fusion analysis and display of multi-source data in rural areas, the method including: collecting basic information, industry-related information and cultural and geographical information of target villages, and constructing a structured multi-source dataset of villages; The historical question data from the village information query website is obtained, and the historical question data is divided into multiple question types according to the query content; Each historical question is associated with its corresponding answer and labeled, and multiple question-and-answer sub-libraries are built according to the question type to form a categorized retrieval-enhanced knowledge base; A village question-answering generative model is constructed based on the village multi-source dataset and the Transformer model, and the model is trained using data from the retrieval enhancement knowledge base. The trained village question-and-answer generative model was optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

[0006] In one feasible implementation, the basic information includes one or more of the following: village name, administrative division, geographical location, population size, population composition, and population distribution; the industry association information includes one or more of the following: industrial chain cooperation relationship data, labor input and output relationship data, natural resource input and output relationship data, and processed product input and output relationship data; the cultural geography information includes one or more of the following: major natural landscape information, major cultural landscape information, and intangible cultural products and crafts information; the information is collected through public channels. The basic information is stored in text format; the industry association information is stored in a directed graph structure, where nodes represent associated entities and directed edges represent the flow direction and association relationship of elements; and the cultural geography information is stored in a multimodal format combining images, videos, and text.

[0007] In one feasible implementation, the construction of the structured village multi-source dataset specifically includes: Using the unique identifier of a village as the core key, the basic information is associated with the corresponding village entity to construct a basic attribute table; Based on the industry association information, each village node is aligned with the village entities in the basic attribute table, and spatial matching is performed based on geographical coordinates to form a village knowledge graph that integrates attributes and relationships. The cultural and geographical information is associated with the village entities through geographical tags and timestamps, and the image and video content features are stored in a vector database and semantically linked with the text description. The data structure of the structured village multi-source dataset is defined as a set V = {V n}; where V represents the set of villages, V n Represents a single village object; Among them, a single village object V n ={B,I,C} contains three attribute sets: B, I, and C. B is the set of basic information attributes, I is the set of industry-related information attributes, and C is the set of cultural and geographical information attributes.

[0008] In one feasible implementation, historical question data from a village information query website is obtained, and the historical question data is divided according to the query content to obtain multiple question types, specifically including: Obtain the village information query website corresponding to each village object in the village set V; The village information query website collects historical question data for a preset period of time, and classifies the questions in the historical question data into three types: B, I, and C, which correspond to the three types of query needs: basic village information query, industry information query, and cultural information query, respectively.

[0009] In one feasible implementation, each historical question is associated and labeled with its corresponding answer, and multiple question-answer sub-bases are constructed according to the question type to form a categorized retrieval-enhanced knowledge base, specifically including: The questions and answers in each historical question data are associated to form QA data pairs, and the question types of the QA data pairs are labeled. Based on the question type, the corresponding QA data pairs are stored in the corresponding question-and-answer sub-libraries to form the categorized retrieval enhancement knowledge base; wherein, the question-and-answer sub-libraries include at least a basic information question-and-answer sub-library, an industry information question-and-answer sub-library, and a cultural information question-and-answer sub-library.

[0010] In one feasible implementation, a village question-answering generative model is constructed based on the village multi-source dataset and the Transformer model, and the model is trained using data from the retrieval enhancement knowledge base. Specifically, this includes: The village multimodal dataset is converted into vector data, stored uniformly in a vectorized knowledge base, and an index retrieval interface is established. A retrieval enhancement module is introduced on the basis of the Transformer decoder, and the retrieval enhancement module is associated with the index retrieval interface to form the village question-answering generative model. The village question-answering generative model is trained by classifying and training the QA data pairs corresponding to each question type in the enhanced knowledge base.

[0011] In one feasible implementation, after constructing a village question-answering generative model based on the village multi-source dataset and the Transformer model, and training the model using data from the retrieval-enhanced knowledge base, the method further includes: Semantic encoding is performed on the target question input by the user into the village question-answering generative model. The data fragment most relevant to the target question is retrieved from the vectorized knowledge base and the data fragment is concatenated as context into the input sequence. The input sequence is fed into the Transformer model to generate a coherent text answer and display it to the user. When the target question is identified as belonging to the category of cultural and geographical information queries, the corresponding multimodal materials of cultural and geographical information are displayed synchronously while the text answer is output; wherein, the multimodal materials include one or more of image materials and video materials.

[0012] In one feasible implementation, the trained village question-answering generative model is optimized using user feedback to obtain the final village multi-source data fusion analysis model, specifically including: Collect feedback data after each user asks a question. After collecting a preset amount of feedback data, train a preset reward model using the collected feedback dataset to learn the mapping relationship between human preferences and answer quality through the reward model. The village question-answering generative model is used as the policy to be optimized, and the output of the reward model is used as the reward signal for reinforcement learning. The model parameters are iteratively updated through the proximal policy optimization algorithm to obtain the optimized village question-answering generative model.

[0013] On the other hand, embodiments of the present invention also provide a rural multi-source data fusion analysis and display system, the system comprising: The data acquisition module is used to collect basic information, industry-related information, and cultural and geographical information of the target villages, and to construct a structured multi-source dataset of the villages. The knowledge base construction module is used to obtain historical question data from village information query websites, and divide the historical question data into multiple question types according to the query content; it associates and annotates each historical question data with the corresponding answer, and constructs multiple question and answer sub-libraries according to the question type to form a classified retrieval enhancement knowledge base; The model training and optimization module is used to construct a village question-answering generative model based on the village multi-source dataset and the Transformer model, and to train the model using data from the retrieval enhancement knowledge base; the trained village question-answering generative model is optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

[0014] Finally, this embodiment of the invention also provides a storage medium, which is a non-volatile computer-readable storage medium storing at least one program, each program including instructions, which, when executed by a terminal, cause the terminal to execute the rural multi-source data fusion analysis and display method.

[0015] Compared with existing technologies, the rural multi-source data fusion analysis and display method, system, and medium provided in this invention have the following beneficial effects: This invention addresses the pain points of scattered, heterogeneous, and difficult-to-use data in rural scenarios by adopting differentiated preprocessing methods such as text storage, directed graph structure storage, and multimodal storage for three different types of data: basic village information, industry-related information, and cultural and geographical information. It establishes a unified structured representation system for village data, solves the pain points of scattered, heterogeneous, and difficult-to-use data in rural scenarios, and provides a standardized and complete data foundation for subsequent intelligent question answering and display.

[0016] This invention constructs a categorized retrieval-enhanced QA knowledge base based on three major categories: basic information, industry information, and cultural information. This enables rapid identification of question types and targeted recall of corresponding answers, avoids information redundancy and matching bias caused by cross-category retrieval, significantly shortens the retrieval path, and improves the accuracy and response speed of question matching.

[0017] A village information intelligent body has been constructed, integrating multi-source data fusion, knowledge retrieval, intelligent generation, and multi-modal display. Users can obtain multi-dimensional information such as basic village overview, industrial structure, and cultural resources in one stop by asking questions in natural language, which reduces the threshold for village information access and improves the inclusiveness and convenience of smart village public information services. Attached Figure Description

[0018] 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 recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart of a method for fusion analysis and display of multi-source data in rural areas provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a rural multi-source data fusion analysis and display system provided in an embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0020] This invention provides a method for the fusion analysis and display of multi-source data in rural areas, such as... Figure 1 As shown, the method for fusion analysis and visualization of multi-source data in rural areas specifically includes steps S101-S105: S101. Collect basic information, industry-related information, and cultural and geographical information of the target villages to construct a structured multi-source dataset of villages.

[0021] Specifically, basic information includes one or more of the following: village name, administrative division, geographical location, population size, population composition, and population distribution; industry-related information includes one or more of the following: data on industrial chain cooperation, labor input and output, natural resource input and output, and processed product input and output; cultural and geographical information includes one or more of the following: information on major natural landscapes, information on major cultural landscapes, and information on intangible cultural products and crafts; and the information is collected through public channels.

[0022] Basic information is stored in text format; industry association information is stored in a directed graph structure, where nodes represent associated entities and directed edges represent the flow direction and association relationship of elements; cultural and geographical information is stored in a multimodal format combining images, videos, and text.

[0023] It should be noted that the basic information, industry-related information, and cultural and geographical information described in this invention were all obtained through legal and publicly available channels. Furthermore, open-source map data was used in conjunction with its open database license agreement to ensure the compliance of information sources. For sensitive location data that could not be directly obtained publicly, access was obtained by applying for dedicated line access from the relevant authorities or by purchasing commercial licenses. All data underwent legality verification and anonymization before use, ensuring no personal privacy was involved, thereby guaranteeing the legality and security of information acquisition and subsequent processing.

[0024] Furthermore, a structured multi-source dataset of villages is constructed, specifically including: Using the unique identifier of each village as the core key, basic information is associated with the corresponding village entity to construct a basic attribute table. Based on industry association information, each village node is aligned with the village entity in the basic attribute table, and spatial matching is performed based on geographical coordinates to form a village knowledge graph that integrates attributes and relationships.

[0025] Cultural geographic information is linked to village entities through geographic tags and timestamps, and image and video content features are stored in a vector database and semantically linked with text descriptions.

[0026] The data structure of the structured village multi-source dataset is defined as a set V = {V n}; where V represents the set of villages, V n Represents a single village object; where the single village object V n={B,I,C} contains three attribute sets: B, I, and C. B is the set of basic information attributes, I is the set of industry-related information attributes, and C is the set of cultural and geographical information attributes.

[0027] S102. Obtain historical question data from the village information query website, and divide the historical question data into multiple question types based on the query content.

[0028] Specifically, the system retrieves the village information query websites corresponding to each village in the village set V. Historical query data for a preset time period is collected from these websites, and the questions are manually labeled and categorized into three types: B, I, and C, corresponding to three types of query needs: basic village information query, industry information query, and cultural information query, respectively.

[0029] S103. Associate and label each historical question with its corresponding answer, and construct multiple question-and-answer sub-databases according to the question type to form a categorized retrieval-enhanced knowledge base.

[0030] Specifically, the questions and answers in each historical question data are associated to form QA data pairs, and the question types of the QA data pairs are labeled.

[0031] Furthermore, based on the question type, the corresponding QA data pairs are stored in the corresponding question-and-answer sub-libraries to form a categorized retrieval-enhanced knowledge base; among them, the question-and-answer sub-libraries include at least a basic information question-and-answer sub-library, an industry information question-and-answer sub-library, and a cultural information question-and-answer sub-library.

[0032] S104. Construct a village question-answering generative model based on village multi-source datasets and the Transformer model, and train the model by retrieving data from the enhanced knowledge base.

[0033] Specifically, the village multimodal dataset is converted into vector data, stored uniformly in a vectorized knowledge base, and an index retrieval interface is established.

[0034] Furthermore, a retrieval enhancement module is introduced based on the Transformer decoder, and this module is associated with the index retrieval interface to form a village question-answering generative model. The village question-answering generative model is then trained by classifying and training the QA data pairs corresponding to each question type in the retrieval enhancement knowledge base.

[0035] Furthermore, the target question input by the user into the village question-answering generative model is semantically encoded. The most relevant data fragments to the target question are retrieved from the vectorized knowledge base and concatenated into the input sequence as context. The input sequence is then fed into the Transformer model to generate a coherent text answer, which is then displayed to the user.

[0036] As a feasible implementation method, a retrieval-enhanced generative question-answering model for village multi-source knowledge bases is constructed based on a pre-trained Transformer model. First, pre-processed and fused basic village information, industry relationship graphs, and cultural multimodal data are uniformly stored in a vectorized knowledge base, and an efficient index retrieval interface is established. Then, a retrieval enhancement module is introduced based on the Transformer decoder: when a user asks a question about a village, such as "the industrial chain cooperation relationships of a certain village," the system first semantically encodes the question, retrieves the most relevant text fragments, graph structures, or multimedia descriptions from the knowledge base, and concatenates these retrieval results as context into the input sequence. Finally, leveraging the autoregressive generative capability of the Transformer and combining it with retrieved evidence information, a coherent and accurate answer is generated word-by-word. This model can dynamically utilize external knowledge during the answering process, ensuring that the answer originates from real village data.

[0037] As a feasible implementation method, when the target question is identified as belonging to the category of cultural and geographical information query, the corresponding multimodal materials of cultural and geographical information are called up and displayed synchronously while outputting the text answer; among them, the multimodal materials include one or more of image materials and video materials.

[0038] In one embodiment, users check in at the village multi-source data fusion analysis and display system and ask questions according to their own needs, such as, "What is the basic situation of XX village? If we are going to this place from Shanghai next Sunday, what is the best way to get there?" The intelligent agent analyzes the question type, distinguishing between basic information query questions and cultural and geographical information query questions. Based on these two question types, it queries the corresponding information in the corresponding question-and-answer sub-database and generates a coherent and smooth answer.

[0039] S105. The trained village question-and-answer generative model is optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

[0040] Specifically, feedback data is collected after each user asks a question. After collecting a preset amount of feedback data, a preset reward model is trained using the collected feedback dataset to learn the mapping relationship between human preferences and answer quality.

[0041] Furthermore, the village question-answering generative model is used as the policy to be optimized, and the output of the reward model is used as the reward signal for reinforcement learning. The model parameters are iteratively updated through the proximal policy optimization algorithm to obtain the optimized village question-answering generative model.

[0042] In one embodiment, to improve the model's response quality in village scenarios, a human feedback reinforcement learning method is employed for optimization. Specifically, for three question scenarios—basic information, industry information, and cultural information—150 users were selected for each scenario, and satisfaction ratings for each model-generated answer were collected after each user asked a question, resulting in a total of 9,000 human feedback data points. The optimization process consists of three steps: First, a reward model is trained using these rating data to learn the mapping relationship between human preferences and answer quality; second, the pre-trained question-answering model is used as the policy to be optimized, and the output of the reward model is used as the reward signal for reinforcement learning; finally, the proximal policy optimization (PPO) algorithm is used to iteratively update the model parameters, making the model's generated answers more likely to receive high-scoring feedback. Through this process, the model gradually learns to generate more accurate answers that better meet the information needs of the village.

[0043] In addition, embodiments of the present invention also provide a rural multi-source data fusion analysis and display system, such as... Figure 2 As shown, the Rural Multi-Source Data Fusion Analysis and Display System 200 specifically includes: The data acquisition module 210 is used to collect basic information, industry-related information and cultural and geographical information of the target village, and to construct a structured multi-source dataset of the village. The knowledge base construction module 220 is used to obtain historical question data from the village information query website, and divide the historical question data according to the query content to obtain multiple question types; associate and label each historical question data with the corresponding answer, and construct multiple question and answer sub-libraries according to the question type to form a classified retrieval enhancement knowledge base; The model training and optimization module 230 is used to construct a village question-answering generative model based on the village multi-source dataset and the Transformer model, and to train the model using data from the retrieval enhancement knowledge base; to optimize the trained village question-answering generative model based on user feedback, and to obtain the final village multi-source data fusion analysis model.

[0044] Finally, the present invention also provides a storage medium, which is a non-volatile computer-readable storage medium storing at least one program, each program including instructions that, when executed by a terminal, cause the terminal to perform: Collect basic information, industry-related information, and cultural and geographical information of the target villages to construct a structured multi-source dataset of villages; The historical question data from the village information query website is obtained, and the historical question data is divided into multiple question types according to the query content; Each historical question is associated with its corresponding answer and labeled, and multiple question-and-answer sub-libraries are built according to the question type to form a categorized retrieval-enhanced knowledge base; A village question-answering generative model is constructed based on the village multi-source dataset and the Transformer model, and the model is trained using data from the retrieval enhancement knowledge base. The trained village question-and-answer generative model was optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

[0045] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0046] The foregoing has described specific embodiments of the present invention. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0047] The above description is merely an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for fusion analysis and visualization of multi-source data in rural areas, characterized in that, The method includes: Collect basic information, industry-related information, and cultural and geographical information of the target villages to construct a structured multi-source dataset of villages; The historical question data from the village information query website is obtained, and the historical question data is divided into multiple question types according to the query content; Each historical question is associated with its corresponding answer and labeled, and multiple question-and-answer sub-libraries are built according to the question type to form a categorized retrieval-enhanced knowledge base; A village question-answering generative model is constructed based on the village multi-source dataset and the Transformer model, and the model is trained using data from the retrieval enhancement knowledge base. The trained village question-and-answer generative model was optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

2. The method for fusion analysis and display of multi-source rural data according to claim 1, characterized in that, The basic information includes one or more of the following: village name, administrative division, geographical location, population size, population composition, and population distribution; the industry-related information includes one or more of the following: industrial chain cooperation relationship data, labor input and output relationship data, natural resource input and output relationship data, and processed product input and output relationship data; the cultural geography information includes one or more of the following: major natural landscape information, major cultural landscape information, and intangible cultural products and crafts information; the information is collected through public channels. The basic information is stored in text format; the industry association information is stored in a directed graph structure, where nodes represent associated entities and directed edges represent the flow direction and association relationship of elements; and the cultural geography information is stored in a multimodal format combining images, videos, and text.

3. The method for fusion analysis and display of multi-source rural data according to claim 1, characterized in that, The construction of the structured village multi-source dataset specifically includes: Using the unique identifier of a village as the core key, the basic information is associated with the corresponding village entity to construct a basic attribute table; Based on the industry association information, each village node is aligned with the village entities in the basic attribute table, and spatial matching is performed based on geographical coordinates to form a village knowledge graph that integrates attributes and relationships. The cultural and geographical information is associated with the village entities through geographical tags and timestamps, and the image and video content features are stored in a vector database and semantically linked with the text description. The data structure of the structured village multi-source dataset is defined as a set V = {V n }; where V represents the set of villages, V n Represents a single village object; Among them, a single village object V n ={B,I,C} contains three attribute sets: B, I, and C. B is the set of basic information attributes, I is the set of industry-related information attributes, and C is the set of cultural and geographical information attributes.

4. The method for fusion analysis and display of multi-source rural data according to claim 3, characterized in that, Historical question data from village information query websites is obtained, and the historical question data is divided into multiple question types based on the query content, specifically including: Obtain the village information query website corresponding to each village object in the village set V; The village information query website collects historical question data for a preset period of time, and classifies the questions in the historical question data into three types: B, I, and C, which correspond to the three types of query needs: basic village information query, industry information query, and cultural information query, respectively.

5. The method for fusion analysis and display of multi-source rural data according to claim 1, characterized in that, Each historical question is associated and labeled with its corresponding answer, and multiple question-and-answer sub-bases are constructed based on the question type to form a categorized, enhanced retrieval knowledge base, specifically including: The questions and answers in each historical question data are associated to form QA data pairs, and the question types of the QA data pairs are labeled. Based on the question type, the corresponding QA data pairs are stored in the corresponding question-and-answer sub-libraries to form the categorized retrieval enhancement knowledge base; wherein, the question-and-answer sub-libraries include at least a basic information question-and-answer sub-library, an industry information question-and-answer sub-library, and a cultural information question-and-answer sub-library.

6. The method for fusion analysis and display of multi-source rural data according to claim 1, characterized in that, A village question-answering generative model is constructed based on the aforementioned village multi-source dataset and the Transformer model, and the model is trained using data from the aforementioned retrieval enhancement knowledge base. Specifically, this includes: The village multimodal dataset is converted into vector data, stored uniformly in a vectorized knowledge base, and an index retrieval interface is established. A retrieval enhancement module is introduced on the basis of the Transformer decoder, and the retrieval enhancement module is associated with the index retrieval interface to form the village question-answering generative model. The village question-answering generative model is trained by classifying and training the QA data pairs corresponding to each question type in the enhanced knowledge base.

7. The method for fusion analysis and display of multi-source rural data according to claim 6, characterized in that, After constructing a village question-answering generative model based on the village multi-source dataset and the Transformer model, and training the model using data from the retrieval-enhanced knowledge base, the method further includes: Semantic encoding is performed on the target question input by the user into the village question-answering generative model. The data fragment most relevant to the target question is retrieved from the vectorized knowledge base and the data fragment is concatenated as context into the input sequence. The input sequence is fed into the Transformer model to generate a coherent text answer and display it to the user. When the target question is identified as belonging to the category of cultural and geographical information queries, the corresponding multimodal materials of cultural and geographical information are displayed synchronously while the text answer is output; wherein, the multimodal materials include one or more of image materials and video materials.

8. The method for fusion analysis and display of multi-source rural data according to claim 1, characterized in that, The trained village question-and-answer generative model was fine-tuned based on user feedback to obtain the final village multi-source data fusion analysis model, which specifically includes: Collect feedback data after each user asks a question. After collecting a preset amount of feedback data, train a preset reward model using the collected feedback dataset to learn the mapping relationship between human preferences and answer quality through the reward model. The village question-answering generative model is used as the policy to be optimized, and the output of the reward model is used as the reward signal for reinforcement learning. The model parameters are iteratively updated through the proximal policy optimization algorithm to obtain the optimized village question-answering generative model.

9. A rural multi-source data fusion analysis and display system, characterized in that, The system includes: The data acquisition module is used to collect basic information, industry-related information, and cultural and geographical information of the target villages, and to construct a structured multi-source dataset of the villages. The knowledge base construction module is used to obtain historical question data from village information query websites, and divide the historical question data into multiple question types according to the query content; it associates and annotates each historical question data with the corresponding answer, and constructs multiple question and answer sub-libraries according to the question type to form a classified retrieval enhancement knowledge base; The model training and optimization module is used to construct a village question-answering generative model based on the village multi-source dataset and the Transformer model, and to train the model using data from the retrieval enhancement knowledge base; the trained village question-answering generative model is optimized based on user feedback to obtain the final village multi-source data fusion analysis model.

10. A storage medium, characterized in that, The storage medium is a non-volatile computer-readable storage medium that stores at least one program, each program including instructions that, when executed by a terminal, cause the terminal to perform a rural multi-source data fusion analysis and display method according to any one of claims 1-8.

Images