LLM model-based text analysis method and device, medium and equipment

Abstract

The invention relates to the technical field of text processing, in particular to a text analysis method and device based on an LLM model, a medium and equipment, and the method comprises the steps: inputting a target text into a preset LLM model, and obtaining a target knowledge graph of the target text; generating a target task set of a target text based on the target knowledge graph; wherein the target task set comprises a plurality of target tasks; sending each target task to a corresponding target analysis model to obtain a target analysis text of each target task; processing all the target analysis texts to obtain a target report of the target text; compared with a traditional manual or existing AI auxiliary query tool, the method has the advantages that complex technical requirement analysis, task planning and execution processes can be automated, the deep research period is greatly shortened, the conversion mode efficiency is improved, and time and labor are reduced.

Description

Technical Field

[0001] This invention relates to the field of text processing technology, and in particular to a text analysis method, apparatus, medium and device based on an LLM model. Background Technology

[0002] Currently, the commercialization of scientific and technological achievements faces core problems such as low efficiency and inaccurate matching. Traditional models rely on manual operation, resulting in long conversion cycles and high costs. Although AI patent search tools can assist in information retrieval and analysis, their function is limited to passive support and cannot achieve automated planning and execution of complex conversion processes. While large language models (LLMs) perform well in language understanding and generation, they still lack automatic planning and task decomposition capabilities when dealing with complex cross-domain and multimodal technical needs, requiring manual intervention to correct the results. In addition, existing multi-agent systems have significant defects in collaborative efficiency. Centralized methods experience an exponential increase in computational complexity due to the increase in the number of agents, while decentralized methods struggle to handle dynamic collaborative tasks. Although some research has attempted to optimize multi-agent collaboration through dynamic task allocation or deep reinforcement learning, challenges remain in task orchestration, result aggregation, and consistency assurance. On the other hand, industry knowledge bases, as the foundation of AI decision-making, generally suffer from poor data quality, weak semantic connections, and delayed updates. The construction of patent knowledge graphs relies on manual annotation, has a low degree of automation, and is difficult to adapt to the needs of rapid technological iteration. This "shallow governance" makes it difficult for the knowledge base to support the high-precision decision-making of AI systems, increasing the risk of "illusion" and thus affecting the reliability of matching scientific and technological achievements. To address these issues, this invention proposes an intelligent transformation method that integrates large AI models, multi-agent collaboration, and a deeply governed knowledge base. Through large AI models, it achieves intelligent decomposition and planning of technological requirements, optimizes the task allocation and execution mechanism of multi-agent systems, and enhances the data relevance and real-time performance of the knowledge base. This promotes the transformation of scientific and technological achievements from "experience-driven" to "intelligent-driven," providing key technological support for scientific and technological innovation. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a text analysis method based on an LLM model, the method comprising the following steps: S100: Input the target text into the preset LLM model to obtain the target knowledge graph of the target text; S200, Based on the target knowledge graph, generate a target task set for the target text; wherein, the target task set includes several target tasks; S300, each of the target tasks is sent to its corresponding target analysis model to obtain the target analysis text of each target task; S400 processes all target analysis texts to obtain a target report of the target texts.

[0004] According to a second aspect of the present invention, a text analysis device based on an LLM model is provided, the device comprising: The first execution module is used to input the target text into the preset LLM model and obtain the target knowledge graph of the target text; The second execution module is used to generate a target task set for the target text based on the target knowledge graph; wherein the target task set includes several target tasks; The third execution module is used to send each of the target tasks to its corresponding target analysis model to obtain the target analysis text of each target task; The fourth execution module is used to process all target analysis texts and obtain target reports of the target texts.

[0005] According to a third aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or the at least one program being loaded and executed by a processor to implement the above-described text analysis method based on the LLM model.

[0006] According to a fourth aspect of the present invention, an electronic device is provided, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0007] The present invention has at least the following beneficial effects: This invention provides a text analysis method based on an LLM model. The method includes the following steps: inputting target text into a preset LLM model to obtain a target knowledge graph of the target text; generating a target task set of the target text based on the target knowledge graph; wherein the target task set includes several target tasks; sending each target task to its corresponding target analysis model to obtain the target analysis text of each target task; processing all the target analysis texts to obtain a target report of the target text. Compared with traditional manual or existing AI-assisted query tools, this invention can automate the complex technical requirement analysis, task planning and execution process, significantly shorten the "deep research" cycle, improve the efficiency of the transformation model and reduce time and effort consumption, and accelerate the discovery and application of scientific and technological achievements. At the same time, the deeply governed industry knowledge base combined with the semantic understanding of the AI ​​large model and the collaborative execution of multiple agents can perform deeper technical associations and requirement matching, reduce the probability of the AI ​​system producing "illusions", and ensure that the output matching results are not only highly relevant but also highly convertible. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart illustrating a text analysis method based on an LLM model provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a text analysis device based on an LLM model provided in an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] Example 1 like Figure 1 As shown, Embodiment 1 of the present invention provides a text analysis method based on an LLM model, the method comprising the following steps: S100: Input the target text into the preset LLM model to obtain the target knowledge graph of the target text; S200, Based on the target knowledge graph, generate a target task set for the target text; wherein, the target task set includes several target tasks; S300, each of the target tasks is sent to its corresponding target analysis model to obtain the target analysis text of each target task; S400 processes all target analysis texts to obtain a target report of the target texts.

[0012] Specifically, the target text is the text containing the question statement input by the target client.

[0013] Specifically, step S100 also includes the following steps: S101, the target text is input into a preset LLM model to obtain an initial tag set and a keyword set for the target text, wherein the initial tags are intent tags describing the target text; any method for obtaining text tags in the prior art is known to those skilled in the art, and will not be described in detail here; any method for obtaining text keywords in the prior art is known to those skilled in the art, and will not be described in detail here.

[0014] S102, Based on the initial tag set of the target text, determine the target tag of the target text.

[0015] S103, based on the target tags of the target text, determine the first target triplet set of the target text.

[0016] S104, based on the first target triplet of the target text and the keyword set of the target text, generate a second target triplet of the target text, so as to generate a target knowledge graph of the target text according to the second target triplet.

[0017] In one specific embodiment, step S102 further includes the following step: S1021, Obtain the initial tag set A={A1, ..., A2} of the target text. i , ..., A m The priority set corresponding to A is B = {B1, ..., B}. i , ..., B m},,A i B is the i-th initial label of the target text. i It is A i The corresponding priority, i, ranges from 1 to m, where m is the number of initial labels for the target text. The priority is the probability value of the target text determining the initial labels through the preset LLM model. Those skilled in the art are familiar with the methods for determining label probabilities in existing large models, which will not be elaborated here.

[0018] S1022, Input A into the preset LLM model to obtain the first label cluster A of the target text. 0 1 and the second tag cluster A of the target text 0 2, wherein the initial tags of the target text in the first tag cluster and the initial tags of the target text in the second tag cluster are mutually exclusive; this can be understood as: S1023, from A 0 In step 1, determine the A corresponding to the highest priority. 0 1max and A 0 In step 2, the highest priority corresponding to A is determined. 0 2max .

[0019] S1024, when A 0 1max A has a higher priority than A. 0 2max When considering the priority of A, 0 1max As the first tag of the target text.

[0020] S1025, when A 0 1max Its priority is less than A. 0 2max When considering the priority of A, 0 2max As the first tag of the target text.

[0021] S1026, when A 0 1max The priority is equal to A 0 2max When determining the priority, according to A 0 1 and A 0 The analysis results of 2 identified the first tag of the target text.

[0022] S1027, Based on the tag cluster corresponding to the first tag of the target text, determine the target tag of the target text.

[0023] Furthermore, step S1027 also includes the following steps: The priority of all tags in the tag cluster corresponding to the first tag of the target text is analyzed to obtain the first analysis result of the target text; When the first analysis result of the target text meets the preset conditions, the target tag of the target text is determined; When the first analysis result of the target text does not meet the preset conditions, the second tag of the target text is determined; The priority of all tags in the tag cluster corresponding to the second tag of the target text is analyzed to obtain the second analysis result of the target text; When the second analysis result of the target text meets the preset conditions, the target tag of the target text is determined; When the second analysis result of the target text does not meet the preset conditions, the target text is adjusted to the prompt words in the preset LLM model, and the target text is re-entered into the preset LLM model.

[0024] Specifically, in step S103, the first target triple set includes several first target triples. The first target triple is the ontology relation triple corresponding to the target text, that is, the triple includes the first ontology, the second ontology, and the association relationship between the first ontology and the second ontology.

[0025] Specifically, in step S104, the second target triple set of the target text includes several second target triples. The second target triple is the entity relation triple corresponding to the target text, that is, the entity relation triple is generated based on the first target triple combined with the keywords of the target text.

[0026] Specifically, in step S200, the target knowledge graph is input into a preset LLM model to generate a target task set for the target text; wherein, the target task set includes several target tasks, and the target task is a task composed of each triple in the target knowledge graph.

[0027] Specifically, the target analysis model described in step S300 is a model that analyzes a single target task.

[0028] Specifically, in step S400, all target analysis texts are input into a preset LLM model to obtain a target report of the target text.

[0029] Example 2 like Figure 2 As shown, Embodiment 2 of the present invention provides a text analysis device based on an LLM model, the device comprising: S100: Input the target text into the preset LLM model to obtain the target knowledge graph of the target text; S200, Based on the target knowledge graph, generate a target task set for the target text; wherein, the target task set includes several target tasks; S300, each of the target tasks is sent to its corresponding target analysis model to obtain the target analysis text of each target task; S400 processes all target analysis texts to obtain a target report of the target texts.

[0030] Specifically, the target text is the text containing the question statement input by the target client.

[0031] Specifically, step S100 also includes the following steps: S101, the target text is input into a preset LLM model to obtain an initial tag set and a keyword set for the target text, wherein the initial tags are intent tags describing the target text; any method for obtaining text tags in the prior art is known to those skilled in the art, and will not be described in detail here; any method for obtaining text keywords in the prior art is known to those skilled in the art, and will not be described in detail here.

[0032] S102, Based on the initial tag set of the target text, determine the target tag of the target text.

[0033] S103, based on the target tags of the target text, determine the first target triplet set of the target text.

[0034] S104, based on the first target triplet of the target text and the keyword set of the target text, generate a second target triplet of the target text, so as to generate a target knowledge graph of the target text according to the second target triplet.

[0035] In one specific embodiment, step S102 further includes the following step: S1021, Obtain the initial tag set A={A1, ..., A2} of the target text. i , ..., A m The priority set corresponding to A is B = {B1, ..., B}. i , ..., B m},,A i B is the i-th initial label of the target text. i It is A i The corresponding priority, i, ranges from 1 to m, where m is the number of initial labels for the target text. The priority is the probability value of the target text determining the initial labels through the preset LLM model. Those skilled in the art are familiar with the methods for determining label probabilities in existing large models, which will not be elaborated here.

[0036] S1022, Input A into the preset LLM model to obtain the first label cluster A of the target text. 0 1 and the second tag cluster A of the target text 0 2, wherein the initial tags of the target text in the first tag cluster and the initial tags of the target text in the second tag cluster are mutually exclusive; this can be understood as: S1023, from A 0 In step 1, determine the A corresponding to the highest priority. 0 1max and A 0 In step 2, the highest priority corresponding to A is determined. 0 2max .

[0037] S1024, when A 0 1max A has a higher priority than A. 0 2max When considering the priority of A, 0 1max As the first tag of the target text.

[0038] S1025, when A 0 1max Its priority is less than A. 0 2max When considering the priority of A, 0 2max As the first tag of the target text.

[0039] S1026, when A 0 1max The priority is equal to A 0 2max When determining the priority, according to A 0 1 and A 0 The analysis results of 2 identified the first tag of the target text.

[0040] S1027, Based on the tag cluster corresponding to the first tag of the target text, determine the target tag of the target text.

[0041] Furthermore, step S1027 also includes the following steps: The priority of all tags in the tag cluster corresponding to the first tag of the target text is analyzed to obtain the first analysis result of the target text; When the first analysis result of the target text meets the preset conditions, the target tag of the target text is determined; When the first analysis result of the target text does not meet the preset conditions, the second tag of the target text is determined; The priority of all tags in the tag cluster corresponding to the second tag of the target text is analyzed to obtain the second analysis result of the target text; When the second analysis result of the target text meets the preset conditions, the target tag of the target text is determined; When the second analysis result of the target text does not meet the preset conditions, the target text is adjusted to the prompt words in the preset LLM model, and the target text is re-entered into the preset LLM model.

[0042] Specifically, in step S103, the first target triple set includes several first target triples. The first target triple is the ontology relation triple corresponding to the target text, that is, the triple includes the first ontology, the second ontology, and the association relationship between the first ontology and the second ontology.

[0043] Specifically, in step S104, the second target triple set of the target text includes several second target triples. The second target triple is the entity relation triple corresponding to the target text, that is, the entity relation triple is generated based on the first target triple combined with the keywords of the target text.

[0044] Specifically, in step S200, the target knowledge graph is input into a preset LLM model to generate a target task set for the target text; wherein, the target task set includes several target tasks, and the target task is a task composed of each triple in the target knowledge graph.

[0045] Specifically, the target analysis model described in step S300 is a model that analyzes a single target task.

[0046] Specifically, in step S400, all target analysis texts are input into a preset LLM model to obtain a target report of the target text.

[0047] Embodiments of the present invention also provide a non-transitory computer-readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a method in the method embodiments, wherein the at least one instruction or the at least one program is loaded and executed by the processor to implement the LLM-based text analysis provided in the above embodiments.

[0048] Embodiments of the present invention also provide an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0049] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. It should also be understood that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.

Claims

1. A text analysis method based on an LLM model, characterized in that, The method includes the following steps: S100: Input the target text into the preset LLM model to obtain the target knowledge graph of the target text; S200, Based on the target knowledge graph, generate a target task set for the target text; wherein, the target task set includes several target tasks; S300, each of the target tasks is sent to its corresponding target analysis model to obtain the target analysis text of each target task; S400 processes all target analysis texts to obtain target reports for the target texts.

2. The text analysis method based on the LLM model according to claim 1, characterized in that, The target text is the text containing the question statement input by the target client.

3. The text analysis method based on the LLM model according to claim 1, characterized in that, Step S100 also includes the following steps: S101, Input the target text into the preset LLM model to obtain the initial tag set and keyword set of the target text; S102, Based on the initial tag set of the target text, determine the target tags of the target text; S103, Based on the target tags of the target text, determine the first target triplet set of the target text; S104, based on the first target triplet of the target text and the keyword set of the target text, generate a second target triplet of the target text, so as to generate a target knowledge graph of the target text according to the second target triplet.

4. The text analysis method based on the LLM model according to claim 3, characterized in that, The initial label is an intent label that describes the target text.

5. A text analysis device based on an LLM model, characterized in that, The device includes: The first execution module is used to input the target text into the preset LLM model and obtain the target knowledge graph of the target text; The second execution module is used to generate a target task set for the target text based on the target knowledge graph; wherein the target task set includes several target tasks; The third execution module is used to send each target task to its corresponding target analysis model to obtain the target analysis text of each target task. The fourth execution module is used to process all target analysis texts and obtain target reports of the target texts.

6. The text analysis device based on the LLM model according to claim 1, characterized in that, The target text is the text containing the question statement input by the target client.

7. The text analysis device based on the LLM model according to claim 1, characterized in that, The first execution module includes: The first acquisition module is used to input the target text into the preset LLM model to obtain the initial label set and keyword set of the target text; The first determining module is used to determine the target tags of the target text based on the initial tag set of the target text; The second determining module is used to determine the first target triplet set of the target text based on the target tags of the target text; The second acquisition module is used to generate a second target triplet of the target text based on the first target triplet and the keyword set of the target text, so as to generate a target knowledge graph of the target text based on the second target triplet.

8. The text analysis method based on the LLM model according to claim 7, characterized in that, The initial label is an intent label that describes the target text.

9. A non-transitory computer-readable storage medium, wherein the storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the text analysis method based on the LLM model as described in any one of claims 1-4.

10. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 8.

Images