Agricultural knowledge service system, method and equipment based on large language model, and medium

Abstract

The invention discloses an agricultural knowledge service system, method and equipment based on a large language model, and a medium, mainly relates to the technical field of agricultural knowledge service, and is used for solving the problems that agricultural information may be uneven, wrong or outdated content exists, and the service efficiency is high in the existing agricultural knowledge acquisition mode. The accuracy of knowledge services is influenced, the coverage of required agricultural fields and subdivided topics cannot be ensured, and diversified requirements are difficult to meet. Comprising the steps of obtaining multi-source agricultural knowledge according to a preset multi-source data obtaining terminal; issuing the multi-source agricultural knowledge to a corresponding expert terminal to obtain a feedback result; identifying a relationship between entity data and entities in the multi-source agricultural knowledge, and labeling the relationship between the entity data and the entities to obtain a trained large language model; and acquiring farmland real-time data or agricultural consultation data, and inputting the farmland real-time data or the agricultural consultation data into the trained large language model to obtain corresponding multi-source agricultural knowledge.

Description

Technical Field

[0001] The present application relates to the technical field of agricultural knowledge services, and in particular to an agricultural knowledge service system, method, device and medium based on a large language model. Background Art

[0002] With the development of modern agricultural technology, agricultural production efficiency continues to improve, and farmers' demand for new technologies and new knowledge is also growing. However, traditional agricultural technology promotion methods often fail to meet farmers' personalized knowledge needs, and there are problems such as slow knowledge update and limited coverage.

[0003] In recent years, agricultural knowledge service solutions mainly involve crawling agricultural information data sources to obtain heterogeneous knowledge information of target agricultural knowledge users; inputting heterogeneous knowledge information into agricultural knowledge graph models to obtain interest prediction results output by agricultural knowledge graph models; and pushing knowledge service content that matches the interest prediction results to target agricultural knowledge users. By representing the potential preferences of users with social relationships, the problem of cold start for new users can be avoided, which greatly improves the pertinence and timeliness of agricultural knowledge services.

[0004] However, the above solution mainly crawls agricultural information data sources from websites. The agricultural information on the Internet may be uneven, with errors or outdated content, which affects the accuracy of knowledge services. In addition, the above solution cannot ensure coverage of the required agricultural fields and sub-topics, and it is difficult to meet diverse needs. Summary of the invention

[0005] In response to the above-mentioned deficiencies in the prior art, the present application provides an agricultural knowledge service system, method, device and medium based on a large language model to solve the existing problems of agricultural knowledge acquisition: agricultural information may be uneven, contain erroneous or outdated content, which affects the accuracy of knowledge services, cannot ensure coverage of required agricultural fields and sub-topics, and is difficult to meet diversified needs.

[0006] In a first aspect, the present application provides an agricultural knowledge service system based on a large language model, the system comprising: The knowledge base construction module is used to obtain multi-source agricultural knowledge according to the preset multi-source data acquisition terminal; send the multi-source agricultural knowledge to the corresponding expert terminal to obtain feedback results; store the feedback results as qualified multi-source agricultural knowledge; deduplicate the qualified multi-source agricultural knowledge and unify it into a preset format; identify the entity data and the relationship between entities in the multi-source agricultural knowledge, and mark the entity data and the relationship between entities; send the marked data to the corresponding expert terminal to obtain secondary marked data; the large language model module is used to use the secondary marked data to train the large language model to obtain the trained large language model; the real-time digital processing module is used to obtain real-time farmland data from sensors installed in actual farmland or obtain agricultural consulting data through a preset user interaction interface, and input the real-time farmland data or agricultural consulting data into the trained large language model to obtain the corresponding multi-source agricultural knowledge.

[0007] The agricultural knowledge service system provided by the embodiment of the present application collects agricultural knowledge by presetting a multi-source data acquisition terminal, thereby ensuring the diversity and extensiveness of information. The feedback mechanism of the expert terminal helps to screen and verify the accuracy of knowledge and eliminate erroneous or outdated content. The secondary annotation data further enhances the accuracy of knowledge and provides a high-quality data set for training a large language model. The integration of multi-source agricultural knowledge ensures that the system can cover a wider range of agricultural fields and subdivided topics. The annotation of entity data and the relationship between entities helps to build a more complete and detailed knowledge graph to meet diverse needs. The real-time digital processing module can quickly acquire data from farmland sensors or user interaction interfaces, and immediately input the data into the large language model for processing. The trained large language model can quickly generate corresponding multi-source agricultural knowledge and provide timely knowledge services. The large language model has natural language processing capabilities and knowledge reasoning capabilities, and can provide intelligent knowledge services according to user needs. The data obtained through the user interaction interface can reflect the personalized needs of users, and the system can provide more accurate knowledge recommendations based on this.

[0008] In one implementation of the present application, the system also includes: a digital twin module, connected to the real-time digital processing module, for integrating real-time farmland data into the digital twin model to generate a three-dimensional virtual simulation model.

[0009] In one implementation of the present application, the system also includes: a user interaction module, which is used to obtain user agricultural consultation data through a preset user interaction interface, and display the multi-source agricultural knowledge output by the input trained large language model.

[0010] In one implementation of the present application, the system also includes a data preference collection module, which is used to obtain all agricultural consulting data of the user currently logged in to the system, obtain the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all the agricultural consulting data, and feed back to the preset multi-source data acquisition terminal to obtain the newly added multi-source agricultural knowledge corresponding to the consulting preferences, and update the training of the large language model again through the multi-source agricultural knowledge.

[0011] In one implementation of the present application, the knowledge base construction module includes a data processing unit for removing duplicate data from qualified multi-source agricultural knowledge using a hash function and a Bloom filter, and unifying the deduplicated multi-source agricultural knowledge into a preset text format.

[0012] In a second aspect, the present application provides an agricultural knowledge service method based on a large language model, the method comprising: According to the preset multi-source data acquisition terminal, multi-source agricultural knowledge is obtained; the multi-source agricultural knowledge is sent to the corresponding expert terminal to obtain feedback results, and the feedback results are stored as qualified multi-source agricultural knowledge; the qualified multi-source agricultural knowledge is deduplicated and unified into a preset format; the entity data and the relationship between entities in the multi-source agricultural knowledge are identified, and the entity data and the relationship between entities are labeled; the labeled data is sent to the corresponding expert terminal to obtain secondary labeled data; the large language model is trained using the secondary labeled data to obtain a trained large language model; real-time farmland data is obtained from sensors installed in actual farmland or agricultural consulting data is obtained through a preset user interaction interface, and the real-time farmland data or agricultural consulting data is input into the trained large language model to obtain corresponding multi-source agricultural knowledge.

[0013] In one implementation of the present application, qualified multi-source agricultural knowledge is deduplicated and unified into a preset format, specifically including: Hash functions and Bloom filters are used to remove duplicate data from qualified multi-source agricultural knowledge, and the deduplicated multi-source agricultural knowledge is unified into a preset text format.

[0014] In one implementation of the present application, the method also includes: obtaining all agricultural consulting data of the user currently logged in to the system, obtaining the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all the agricultural consulting data, feeding back to a preset multi-source data acquisition terminal, obtaining new multi-source agricultural knowledge corresponding to the consulting preferences, and updating and training the large language model again through the multi-source agricultural knowledge.

[0015] In a third aspect, the present application provides an agricultural knowledge service device based on a large language model, the device comprising: processor; and a memory having executable codes stored thereon, which, when executed, causes the processor to execute an agricultural knowledge service method based on a large language model as described above.

[0016] In a fourth aspect, the present application provides a non-volatile computer storage medium having computer instructions stored thereon, which when executed implement an agricultural knowledge service method based on a large language model as described above.

[0017] Those skilled in the art can understand that the present application has at least the following beneficial effects: The present application provides an agricultural knowledge service system, method, device and medium based on a large language model, which collects agricultural knowledge through a preset multi-source data acquisition terminal to ensure the diversity and extensiveness of information. The feedback mechanism of the expert terminal helps to screen and verify the accuracy of knowledge, solves the problem that agricultural information may be uneven, contain erroneous or outdated content, and affect the accuracy of knowledge services. The secondary annotation data further enhances the accuracy of knowledge and provides a high-quality data set for training a large language model. The integration of multi-source agricultural knowledge ensures that the system can cover a wider range of agricultural fields and subdivided topics. The annotation of entity data and the relationship between entities helps to build a more complete and detailed knowledge graph to meet diverse needs. The real-time digital processing module can quickly obtain data from farmland sensors or user interaction interfaces, and immediately input the large language model for processing. The trained large language model can quickly generate corresponding multi-source agricultural knowledge and provide timely knowledge services. The large language model has natural language processing capabilities and knowledge reasoning capabilities, and can provide intelligent knowledge services according to user needs. The data obtained through the user interaction interface can reflect the personalized needs of users, and the system can provide more accurate knowledge recommendations based on this. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Some embodiments of the present disclosure are described below with reference to the accompanying drawings, in which: Figure 1 It is a schematic diagram of the internal structure of an agricultural knowledge service system based on a large language model provided in an embodiment of the present application.

[0019] Figure 2 This is a flow chart of an agricultural knowledge service method based on a large language model provided in an embodiment of the present application.

[0020] Figure 3 It is a schematic diagram of the internal structure of an agricultural knowledge service device based on a large language model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] It should be understood by those skilled in the art that the embodiments described below are only preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented through the preferred embodiments. The preferred embodiments are only used to explain the technical principles of the present disclosure, and are not used to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work should still fall within the protection scope of the present disclosure.

[0022] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0023] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0024] This application Figure 1 An agricultural knowledge service system based on a large language model is provided in the embodiment of the present application. Figure 1 As shown, the system provided in the embodiment of the present application mainly includes: The knowledge base construction module 110 is used to obtain multi-source agricultural knowledge according to a preset multi-source data acquisition terminal; send the multi-source agricultural knowledge to the corresponding expert terminal to obtain feedback results; store the feedback results as qualified multi-source agricultural knowledge; deduplicate the qualified multi-source agricultural knowledge and unify it into a preset format; identify the entity data and the relationship between entities in the multi-source agricultural knowledge, and mark the entity data and the relationship between entities; send the marked data to the corresponding expert terminal to obtain secondary marked data.

[0025] It should be noted that, according to the preset multi-source data acquisition terminal, it can correspond to the upload terminal of various agricultural-related literature materials, expert opinions, research reports, and technical manuals.

[0026] Before sending the multi-source agricultural knowledge to the corresponding expert terminal, the present application can perform simple KNN clustering on the multi-source agricultural knowledge to determine the corresponding type, and then send the multi-source agricultural knowledge to the expert terminal corresponding to the type to obtain feedback results.

[0027] The qualified multi-source agricultural knowledge can be deduplicated and unified into a preset format, which can be obtained by the data processing unit in the knowledge base construction module 110 .

[0028] For example, hash functions and Bloom filters are used to remove duplicate data from qualified multi-source agricultural knowledge, and the deduplicated multi-source agricultural knowledge is unified into a preset text format.

[0029] For example: convert all documents into Markdown or HTML format, and further use natural language processing tools for spell checking and grammar correction. By combining automated scripts and manual review, the accuracy and consistency of the data can be verified to ensure the accuracy and reliability of the knowledge base.

[0030] In this application, entity data and relationships between entities in multi-source agricultural knowledge are identified, and entity data and relationships between entities are annotated; the annotated data is sent to the corresponding expert terminal to obtain secondary annotated data, mainly for: A method combining machine automatic labeling and manual labeling is adopted. Named entity recognition (entity data) and relationship (relationship between entities) extraction technology are used to automatically label key entities and relationships in the text, and then manual labeling is performed through crowdsourcing platforms or internal teams (expert terminals) to ensure the accuracy of labeling. Labeling tools are used to improve labeling efficiency and quality, so as to facilitate subsequent model training.

[0031] The large language model module 120 is used to train the large language model using the secondary labeled data to obtain a trained large language model.

[0032] The process of training a large language model can be: Pre-trained model selection: Select the large language model with the best performance as the basic model, evaluate the performance of different models through benchmark tests and actual application scenarios, and select the most suitable model for the agricultural field.

[0033] Fine-tuning training: Divide the data set into training set, validation set and test set, usually 70%, 15% and 15% respectively. Based on the demand analysis of various agricultural scenarios, agricultural technical knowledge is divided into different topics to form two large model downstream tasks: knowledge object recognition and knowledge question answering. Combined with small sample high-quality training corpus, the basic model is fine-tuned to enable it to have the ability to answer complex agricultural problems. Use strategies such as transfer learning, fine-tuning and multi-task learning to optimize model performance.

[0034] Model optimization: Use data enhancement techniques (such as synonym replacement and sentence reorganization) to increase the diversity of training data. Combine the retrieval system and the generative model to assist in generating answers by retrieving relevant document fragments, alleviate the hallucination problem of large models, and improve the semantic similarity and answer accuracy of the model. Build a knowledge graph in the agricultural field, provide structured background knowledge, and assist the model in generating more accurate answers.

[0035] The real-time digital processing module 130 is used to obtain real-time farmland data from sensors installed in actual farmland or obtain agricultural consulting data through a preset user interaction interface, and input the real-time farmland data or agricultural consulting data into a trained large language model to obtain corresponding multi-source agricultural knowledge.

[0036] In addition, the present application can also be displayed based on three-dimensional data.

[0037] The specific process can be: The digital twin module is connected to the real-time digital processing module 130 and is used to integrate the real-time farmland data (growth conditions, climate conditions, soil moisture and other key parameters) into the digital twin model to generate a three-dimensional virtual simulation model.

[0038] In addition, in order to achieve effective data communication, the system also includes: a user interaction module, which is used to obtain user agricultural consulting data through a preset user interaction interface, and display the multi-source agricultural knowledge output by the input trained large language model.

[0039] It should be noted that the user interaction module supports multi-channel access: it supports access to the system through multiple channels such as web pages, mobile applications, social media, etc., so that users can obtain information anytime and anywhere; the user interaction module supports natural language processing: using natural language processing technology, it supports users to ask questions through text input or voice, and the system can understand the user's intentions and give accurate answers; the user interaction module supports multi-round dialogue support: it supports users to conduct multiple rounds of questions and answers, and the system can remember contextual information and provide coherent and relevant answers.

[0040] In addition, in order to adapt to the current user's habits, the system also includes a data preference collection module, which is used to obtain all agricultural consulting data of the user currently logged in to the system, and obtain the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all agricultural consulting data. The module feeds back to the preset multi-source data acquisition terminal to obtain the newly added multi-source agricultural knowledge corresponding to the consulting preferences, and updates the large language model again through the multi-source agricultural knowledge.

[0041] The embodiment provides an agricultural knowledge service method based on a large language model, such as Figure 2 As shown, the method provided in the embodiment of the present application mainly includes the following steps: Step 210: Obtain multi-source agricultural knowledge according to a preset multi-source data acquisition terminal; send the multi-source agricultural knowledge to a corresponding expert terminal, obtain feedback results, and store the feedback results as qualified multi-source agricultural knowledge.

[0042] Step 220, deduplicate qualified multi-source agricultural knowledge and unify it into a preset format; identify entity data and relationships between entities in the multi-source agricultural knowledge, and annotate the entity data and relationships between entities; send the annotated data to the corresponding expert terminal to obtain secondary annotated data.

[0043] Among them, the qualified multi-source agricultural knowledge is deduplicated and unified into a preset format, which can be specifically: Hash functions and Bloom filters are used to remove duplicate data from qualified multi-source agricultural knowledge, and the deduplicated multi-source agricultural knowledge is unified into a preset text format.

[0044] Step 230: Use the secondary labeled data to train the large language model to obtain a trained large language model.

[0045] Step 240: Acquire real-time farmland data from sensors installed in actual farmland or obtain agricultural consulting data through a preset user interaction interface, input the real-time farmland data or agricultural consulting data into the trained large language model to obtain corresponding multi-source agricultural knowledge.

[0046] In some embodiments, the method further comprises: Obtain all agricultural consulting data of the user currently logged in to the system, obtain the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all agricultural consulting data, feed it back to the preset multi-source data acquisition terminal, obtain the newly added multi-source agricultural knowledge corresponding to the consulting preferences, and update the training of the large language model again through the multi-source agricultural knowledge.

[0047] The above is a method embodiment of the present application. Based on the same inventive concept, the present application embodiment also provides an agricultural knowledge service device based on a large language model. Figure 3 As shown, the device includes: a processor; and a memory on which executable codes are stored. When the executable codes are executed, the processor executes an agricultural knowledge service method based on a large language model as described in the above-mentioned embodiment.

[0048] Specifically, the server side obtains multi-source agricultural knowledge according to a preset multi-source data acquisition terminal; sends the multi-source agricultural knowledge to the corresponding expert terminal to obtain feedback results, and stores the feedback results as qualified multi-source agricultural knowledge; deduplicates the qualified multi-source agricultural knowledge and unifies it into a preset format; identifies entity data and relationships between entities in the multi-source agricultural knowledge, and annotates the entity data and relationships between entities; sends the annotated data to the corresponding expert terminal to obtain secondary annotated data; uses the secondary annotated data to train a large language model to obtain a trained large language model; obtains real-time farmland data from sensors installed in actual farmland or obtains agricultural consulting data through a preset user interaction interface, and inputs the real-time farmland data or agricultural consulting data into the trained large language model to obtain corresponding multi-source agricultural knowledge.

[0049] In addition, an embodiment of the present application further provides a non-volatile computer storage medium on which executable instructions are stored. When the executable instructions are executed, an agricultural knowledge service method based on a large language model as described above is implemented.

[0050] So far, the technical solutions of the present disclosure have been described in combination with the above multiple embodiments, but it is easy for those skilled in the art to understand that the protection scope of the present disclosure is not limited to these specific embodiments. Without departing from the technical principles of the present disclosure, those skilled in the art can split and combine the technical solutions in the above-mentioned various embodiments, and can also make equivalent changes or replacements to the relevant technical features. Any changes, equivalent replacements, improvements, etc. made within the technical concept and / or technical principle of the present disclosure will fall within the protection scope of the present disclosure.

Claims

1. An agricultural knowledge service system based on a large language model, characterized in that: The system comprises: The knowledge base construction module is used to obtain multi-source agricultural knowledge according to the preset multi-source data acquisition terminal; send the multi-source agricultural knowledge to the corresponding expert terminal to obtain feedback results; store the feedback results as qualified multi-source agricultural knowledge; perform deduplication processing on the qualified multi-source agricultural knowledge and unify it into a preset format; identify the entity data and the relationship between entities in the multi-source agricultural knowledge, and mark the entity data and the relationship between entities; send the marked data to the corresponding expert terminal to obtain secondary marked data; The large language model module is used to train the large language model using secondary annotation data to obtain a trained large language model; The real-time digital processing module is used to obtain real-time farmland data from sensors installed in actual farmland or obtain agricultural consulting data through a preset user interaction interface, and input the real-time farmland data or agricultural consulting data into the trained large language model to obtain corresponding multi-source agricultural knowledge.

2. The agricultural knowledge service system based on a large language model according to claim 1, characterized in that: The system also includes: a digital twin module, which is connected to the real-time digital processing module and is used to integrate real-time farmland data into the digital twin model to generate a three-dimensional virtual simulation model.

3. The agricultural knowledge service system based on a large language model according to claim 1, characterized in that: The system also includes: a user interaction module, which is used to obtain user agricultural consultation data through a preset user interaction interface, and display the multi-source agricultural knowledge output by the input trained large language model.

4. The agricultural knowledge service system based on a large language model according to claim 1, characterized in that: The system also includes a data preference collection module, which is used to obtain all agricultural consulting data of the user currently logged in to the system, obtain the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all the agricultural consulting data, and feed back to the preset multi-source data acquisition terminal to obtain the newly added multi-source agricultural knowledge corresponding to the consulting preferences, and update the training of the large language model again through the multi-source agricultural knowledge.

5. The agricultural knowledge service system based on a large language model according to claim 1, characterized in that: The knowledge base building module includes a data processing unit, It is used to remove duplicate data from qualified multi-source agricultural knowledge using hash functions and Bloom filters, and unify the deduplicated multi-source agricultural knowledge into a preset text format.

6. An agricultural knowledge service method based on a large language model, characterized in that: The method comprises: According to the preset multi-source data acquisition terminal, multi-source agricultural knowledge is obtained; the multi-source agricultural knowledge is sent to the corresponding expert terminal, feedback results are obtained, and the feedback results are stored as qualified multi-source agricultural knowledge; De-duplicate qualified multi-source agricultural knowledge and unify it into a preset format; identify entity data and relationships between entities in multi-source agricultural knowledge, and annotate entity data and relationships between entities; send the annotated data to the corresponding expert terminal to obtain secondary annotated data; Use the secondary labeled data to train the large language model and obtain a trained large language model; Real-time farmland data can be obtained from sensors installed in actual farmland, or agricultural consulting data can be obtained through a preset user interaction interface. The real-time farmland data or agricultural consulting data can be input into a trained large language model to obtain corresponding multi-source agricultural knowledge.

7. The agricultural knowledge service method based on a large language model according to claim 6, characterized in that: Qualified multi-source agricultural knowledge is deduplicated and unified into a preset format, including: Hash functions and Bloom filters are used to remove duplicate data from qualified multi-source agricultural knowledge, and the deduplicated multi-source agricultural knowledge is unified into a preset text format.

8. The agricultural knowledge service method based on a large language model according to claim 6, characterized in that: The method further comprises: Obtain all agricultural consulting data of the user currently logged in to the system, obtain the consulting preferences of the user currently logged in to the system based on the multi-source agricultural knowledge corresponding to all agricultural consulting data, feed it back to the preset multi-source data acquisition terminal, obtain the newly added multi-source agricultural knowledge corresponding to the consulting preferences, and update the training of the large language model again through the multi-source agricultural knowledge.

9. An agricultural knowledge service device based on a large language model, characterized in that: The device comprises: processor; and a memory having executable codes stored thereon, which, when executed, causes the processor to execute an agricultural knowledge service method based on a large language model as described in any one of claims 6 to 8.

10. A non-volatile computer storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the agricultural knowledge service method based on a large language model as described in any one of claims 6 to 8 is implemented.

Images