Remote sensing long document generation method combining multi-granularity recall and multi-step reflection

Abstract

The invention discloses a remote sensing long document generation method combining multi-granularity recall and multi-step reflection. The remote sensing long document generation method comprises the following steps: S1, constructing a long document knowledge structure set; s2, an expert collaboration model is constructed, the expert collaboration model comprises three intelligent agents, namely, a Planer, an Executor and a Master, and the three intelligent agents are respectively a Planer agent, an Executor agent and a Master agent; s3, starting a planer of the CoE model to execute planning on the whole writing task; s4, analyzing the input data, and if a model essay is input, analyzing the input model essay at the same time; s5, generating a document outline; s6, generating the content of each chapter and paragraph by adopting a multi-granularity RAG technology; s7, merging the paragraph contents to form a complete document; and S8, checking the content continuity of the whole content by the Master, triggering the modification of the related paragraph content if incoherent content is found, and outputting a final result in a front view after the checking is qualified. According to the method, the semantic consistency and the content coherence in the ultra-long text generation process are improved, and the effect is better than that of direct single-time generation.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing large language models, and in particular to a remote sensing long document generation method combining multi-granularity recall and multi-step reflection. Background Art

[0002] The application of generative AI based on the LLM big model in office automation scenarios is becoming more and more mature, and the automatic generation of long documents is one of the typical applications. LLM can efficiently write tens of thousands of words of long articles based on massive reference knowledge, showing performance that exceeds traditional methods and providing great convenience for users. However, in specific fields, especially in the highly professional and knowledge-intensive remote sensing field, the existing big models still face many challenges in generating long documents.

[0003] First, existing large models are often limited by the length of generation when generating long documents. On the one hand, this is related to the architectural design of the LLM series of models. Although the Transformer architecture has advantages in processing long sequence data, its computational complexity increases quadratically with the length of the sequence, which makes it difficult to support the efficient generation of ultra-long texts of more than 10,000 words in practical applications. On the other hand, the "memory decay" phenomenon is prone to occur in the process of generating long texts, that is, the model's memory of early input information gradually weakens, affecting the coherence and logic of the generated content. In addition, in order to ensure generation efficiency, a maximum generation length threshold is set in many application scenarios, which further limits the length of the document.

[0004] Secondly, even within an acceptable length range, the quality of text generated by large models still has room for improvement. This is reflected in the following aspects:

[0005] 1) Factual errors: For professional fields that require a high degree of accuracy, such as remote sensing, which has higher requirements for the professionalism of text content, professional knowledge errors may appear in the text generated by the model, affecting the authority and credibility of the document.

[0006] 2) Logical incoherence: Long documents usually contain multiple chapters or paragraphs, which require good connections between the contents of each part. However, in the actual generation process, the logical relationship between different parts may not be tight enough, or even contradictory, which reduces the reading experience.

[0007] 3) How to generate full-text content that is consistent with the outline according to the document outline structure specified by the user is another urgent problem to be solved. Ideally, the user can first provide a detailed outline, including the topics, key points and other information of each chapter, and then the model will automatically generate a complete document based on the outline. However, the current level of technology is still difficult to fully achieve this goal. On the one hand, the model needs to have strong understanding ability to accurately parse the instructions in the outline; on the other hand, it also needs to have excellent planning ability to ensure that the generated content strictly follows the outline framework while maintaining the richness and creativity of the content. In fact, although many generated results roughly meet the outline requirements, they are often unsatisfactory in the details, such as deviating from the topic and omitting important information.

[0008] In summary, although the technology of long document generation based on large models has made certain progress, there are still problems such as limited generation length, poor generation quality, and poor consistency between the outline and the overall content. These problems not only restrict the application scope of the technology, but also point out the direction for further research. The improvement direction focuses on optimizing the model architecture, improving the generation quality, and enhancing the consistent understanding and execution capabilities of complex long document generation tasks, so as to promote the application of this field in a more mature and stable direction. Summary of the invention

[0009] In order to solve the existing problems, the present invention provides a remote sensing long document generation method combining multi-granularity recall and multi-step reflection. The specific scheme is as follows:

[0010] A remote sensing long document generation method combining multi-granularity recall and multi-step reflection includes the following steps:

[0011] S1, construct a long document knowledge structure set;

[0012] S2, build an expert collaboration model, namely the CoE model, which includes three agents, namely Planer, Executor, and Master. Among them, Planer is responsible for overall planning, Executor is responsible for execution, and the Master agent is responsible for result summary and verification;

[0013] S3, after the user inputs the writing request, the Planner of the CoE model starts to plan the overall writing task execution;

[0014] S4, parsing the input data, and if there is a model essay input, parsing the input model essay at the same time;

[0015] S5, generating a document outline based on the parsed data and sample papers;

[0016] S6, based on the outline, the system uses multi-granularity RAG technology to extract relevant information from the knowledge base and the super-long context to generate the content of each chapter and paragraph;

[0017] S7, merging the generated chapter and paragraph contents to form a complete document;

[0018] S8, after the complete document content is generated, the Master in the CoE model will check the content coherence of the entire content. If incoherent content is found, the modification of the content of the relevant paragraphs will be triggered until the final result is output after the inspection is qualified.

[0019] Preferably, step S1 is constructed as follows:

[0020] S11, study long document examples;

[0021] S12, structurally splitting the template input by the user, extracting the outline structure and key information of each paragraph, and forming a long document writing template, wherein the key information of each paragraph includes the main body and keywords.

[0022] Preferably, the specific steps of learning in step S11 include:

[0023] S111, performing structural analysis on the long document sample, extracting the title, outline, chapter result, key information of each paragraph, and full text of each paragraph corresponding to each sample, wherein the key information of each paragraph includes the theme and keywords of the paragraph, and forming writing knowledge data;

[0024] S112, constructing the index structure. Specifically, according to the characteristics of long documents, designing a variety of segmentation granularities, keyword extraction and different indexing methods. The different granularities include paragraph, chapter and full text granularities. The different indexing methods include keyword index, title vector index and chapter index.

[0025] Preferably, the base model generated by the content in steps S5 and S6 is a domain base model trained with remote sensing knowledge.

[0026] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. After the computer program is run, any of the above methods is executed.

[0027] The present invention also discloses a computer system, including a processor and a storage medium, wherein a computer program is stored on the storage medium, and the processor reads and runs the computer program from the storage medium to execute any of the methods described above.

[0028] The beneficial effects of the present invention are:

[0029] The effect of the present invention is to achieve the automatic and efficient generation of high-quality long documents in the field of remote sensing. Through the pre-built domain knowledge base and database, the system can understand the topic input by the user and automatically perform steps such as outline formulation and chapter writing, which greatly improves the speed and efficiency of document generation. The following are the main functional features of the present invention:

[0030] (1) Automated process: The user only needs to provide a writing topic, and the present invention can automatically execute the entire process from outline preparation to article writing.

[0031] (2) Reference model essay imitation mechanism: Users can specify the structure of model essays, and the present invention will imitate these structures, thereby improving the personalization and professionalism of the document.

[0032] (3) Domain base model: The present invention adopts a domain base model trained with remote sensing knowledge to ensure the professionalism and accuracy of the generated content.

[0033] (4) Multi-granularity RAG technology: In the paragraph generation stage, the system uses multi-granularity RAG technology to effectively extract relevant information from the knowledge base and ultra-long context, thereby enhancing the coherence of the content.

[0034] (5) CoE multi-expert collaboration model: This invention simulates the multi-expert collaboration model and realizes accurate task processing through the collaborative work of three intelligent agents: Planner, Executor, and Master. The Master agent is responsible for checking the coherence of the document and triggering content modification when necessary to ensure the quality of the document.

[0035] (6) High professionalism: Since the present invention uses a remote sensing field base model in the S5 and S6 stages, it is superior to general large-scale language models in terms of the professionalism of remote sensing content generation.

[0036] In summary, the present invention realizes the efficient, professional and coherent generation of long documents in the remote sensing field through automated processes, imitation writing mechanisms, domain expertise bases, multi-granularity RAG technology and multi-expert collaborative models. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 is a flowchart of the method of the present invention;

[0039] Figure 2 This is an example diagram for generating a long document in the remote sensing field in the embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] The present invention is a long document generation model solution in the remote sensing field that combines multi-granularity RAG retrieval recall with multi-step reflection. Multi-granularity RAG retrieval is used for semantic understanding and retrieval of long contexts to improve semantic consistency in the process of generating ultra-long texts. When multi-step reflection is applied to the outline planning of long texts and the generation of content for each chapter, the content coherence of each chapter and the overall outline can be confirmed through multi-step reflection, which is better than the effect of direct single generation.

[0042] like Figure 1 ,A remote sensing long document generation method combining multi-granularity recall and multi-step reflection, comprising the following steps:

[0043] S1, construct a long document knowledge structure set. The long document knowledge structure set in the remote sensing field is to support the generation of long documents in the remote sensing field. The pre-constructed domain long document chapter structure knowledge base and database. Usually, the user only needs to provide a writing topic and the system can automatically start a series of overall processes such as outline formulation and chapter writing. In particular, the present invention adds a reference model writing mechanism, and the user can directly specify the structure of the model and request a large model to be written.

[0044] The specific steps are as follows:

[0045] S11, study long document examples.

[0046] The specific steps of learning include:

[0047] S111, perform structural analysis on long document examples, extract the title, outline, chapter results, key information of each paragraph, and full text of each paragraph corresponding to each example, wherein the key information of each paragraph includes the paragraph theme and keywords, to form writing knowledge data.

[0048] S112, constructing the index structure. Specifically, according to the characteristics of long documents, designing a variety of segmentation granularities, keyword extraction and different indexing methods. The different granularities include paragraph, chapter and full text granularities. The different indexing methods include keyword index, title vector index and chapter index.

[0049] S12, structurally splitting the template input by the user, extracting the outline structure and key information of each paragraph, and forming a long document writing template, wherein the key information of each paragraph includes the main body and keywords.

[0050] S2, build an expert collaboration model, namely the CoE (Collaboration-of-Experts) model, which refers to a multi-expert collaboration model, in which multiple agents play multiple expert roles in the writing process. The core concept of the CoE expert collaboration model in the large model is to achieve accurate task processing through the collaborative work of multiple expert models. The CoE model of the present invention includes three agents, namely Planer, Executor, and Master, among which Planer is responsible for overall planning, Executor is responsible for execution, and the Master supervisor agent is responsible for result summary and verification.

[0051] S3: After the user inputs the writing request, the Planner of the CoE model starts planning the overall writing task, determining the general framework of the document and the steps to be followed.

[0052] S4, parse the input data, and if there is a sample text input, parse the input sample text at the same time. This step is the key to ensure that the document content is consistent with user needs and given data.

[0053] S5, based on the parsed data and sample papers, generate a document outline. This step is the basis of document writing and ensures the logic and completeness of the document structure.

[0054] S6, based on the outline, the system uses multi-grain RAG (multi-grain Retrieval-augmented Generation) technology to extract relevant information from the knowledge base and the ultra-long context to generate the content of each chapter and paragraph. This step ensures the professionalism and coherence of the generated content. Multi-grain RAG technology can effectively extract relevant information from the knowledge base and the ultra-long context, further ensuring the professionalism and coherence of the generated content. This technology can better understand and utilize contextual information by extracting information from different granularities (such as words, phrases, sentences, etc.), thereby improving the quality and relevance of the generated text.

[0055] Among them, the base model generated by the content in steps S5 and S6 is the domain base model GisRS-LongWriter trained with remote sensing knowledge.

[0056] S7, merge the generated chapter and paragraph contents to form a complete document. This step involves adjusting and optimizing the logical relationship between paragraphs to ensure the fluency and consistency of the document.

[0057] S8, after the complete document content is generated, the intelligent agent Master in the CoE model checks the content coherence of the entire content. If incoherent content is found, the modification of the content of the relevant paragraphs is triggered until the final result is output after the inspection is qualified.

[0058] The present invention realizes the automatic and efficient generation of high-quality long documents in the field of remote sensing. Through the pre-built domain knowledge base and database, the system can understand the topic input by the user and automatically perform steps such as outline formulation and chapter writing, greatly improving the speed and efficiency of document generation.

[0059] The main functional features of the present invention are as follows:

[0060] (1) Automated process: The user only needs to provide a writing topic, and the present invention can automatically execute the entire process from outline preparation to article writing.

[0061] (2) Reference model essay imitation mechanism: Users can specify the structure of model essays, and the present invention will imitate these structures, thereby improving the personalization and professionalism of the document.

[0062] (3) Domain base model: The present invention adopts a domain base model trained with remote sensing knowledge, trains a domain long document fine-tuned base model, GisRS-LongWriter, and expands the single input and output length of LLM to more than 10,000 words, ensuring the professionalism and accuracy of the generated content and improving the quality of domain long text generation.

[0063] (4) Multi-grain RAG technology: Multi-grain RAG (multi-grain Retrieval-augmented Generation) related content recall is introduced to improve the ability to recall key information in long text retrieval, improve the semantic understanding and retrieval ability of long contexts, and improve the semantic consistency in the long text output process.

[0064] (5) CoE multi-expert collaboration model: The CoE (Collaboration-of-Experts) thinking chain of multi-agent writing is introduced to simulate the multi-expert collaboration mode. When planning the outline of a long article and generating the content of each chapter, it can go through multiple steps of reflection to confirm that the content of each chapter is consistent with the overall outline. After the overall writing is completed, the master agent will conduct feedback inspection to ensure the content consistency of the overall output results. Specifically, the task is accurately processed through the collaborative work of the three intelligent agents Planner, Executor and Master. The Master intelligent agent is responsible for checking the consistency of the document and triggering content modification when necessary to ensure the quality of the document.

[0065] (6) High professionalism: Since the present invention uses a remote sensing field base model in the S5 and S6 stages, it is superior to general large-scale language models in terms of the professionalism of remote sensing content generation.

[0066] In summary, the present invention realizes the efficient, professional and coherent generation of long documents in the remote sensing field through automated processes, imitation writing mechanisms, domain expertise bases, multi-granularity RAG technology and multi-expert collaborative models.

[0067] The embodiments of the present invention can be seen Figure 2 The data examples shown in Table 1 are representative of general document writing. In general, users only provide the writing topic, and the system automatically starts a series of overall processes such as outline formulation and chapter writing. The example 2 model writing listed in Table 1 is a reference model writing mechanism designed by this system. Users can directly specify the structure of the model and request large-scale model writing. Any super-long text (>10,000 words) actually written by the user can be used as a template or example for long text generation.

[0068] The present invention also discloses a computer-readable storage medium and a computer system, wherein a computer program is stored on the computer-readable storage medium, and after the computer program is run, any of the above methods is executed. A computer system includes a processor and a storage medium, wherein the computer program is stored on the storage medium, and the processor reads and runs the computer program from the storage medium to execute any of the above methods.

[0069] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0070] The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0071] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0072] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, a server, or other remote source using a coaxial cable, a fiber optic cable, a twisted pair, a digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, wherein disk often reproduces data magnetically, while disc reproduces data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0073] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

[0074] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote sensing long document generation method combining multi-granularity recall and multi-step reflection, characterized in that: The following steps are involved: S1, construct a set of long document knowledge structures; S2, build an expert collaboration model, namely the CoE model, which includes three agents, namely Planer, Executor, and Master. Among them, Planer is responsible for overall planning, Executor is responsible for execution, and the Master agent is responsible for result summary and verification; S3, after the user inputs the writing request, the Planner of the CoE model starts to plan the overall writing task execution; S4, parsing the input data, and if there is a model essay input, parsing the input model essay at the same time; S5, generating a document outline based on the parsed data and sample papers; S6, based on the outline, uses multi-granularity RAG technology to extract relevant information from the knowledge base and the super-long context to generate the content of each chapter and paragraph; S7, merging the generated chapter and paragraph contents to form a complete document; S8, after the complete document content is generated, the Master in the CoE model will check the content coherence of the entire content. If incoherent content is found, the modification of the content of the relevant paragraphs will be triggered until the final result is output after the inspection is qualified.

2. The method according to claim 1, characterized in that: Step S1 builds the steps as follows: S11, study long document examples; S12, structurally splitting the template input by the user, extracting the outline structure and key information of each paragraph, and forming a long document writing template, wherein the key information of each paragraph includes the main body and keywords.

3. The method according to claim 1, characterized in that: The specific steps learned in step S11 include: S111, performing structural analysis on the long document sample, extracting the title, outline, chapter result, key information of each paragraph, and full text of each paragraph corresponding to each sample, wherein the key information of each paragraph includes the theme and keywords of the paragraph, and forming writing knowledge data; S112, constructing the index structure. Specifically, according to the characteristics of long documents, designing a variety of segmentation granularities, keyword extraction and different indexing methods. The different granularities include paragraph, chapter and full text granularities. The different indexing methods include keyword index, title vector index and chapter index.

4. The method according to claim 1, characterized in that: The base model generated by the content in steps S5 and S6 is a domain base model trained with remote sensing knowledge.

5. A computer-readable storage medium, characterized in that: A computer program is stored on the medium, and after the computer program is run, the method according to any one of claims 1 to 4 is executed.

6. A computer system, characterized in that: The method comprises a processor and a storage medium, wherein a computer program is stored in the storage medium, and the processor reads and runs the computer program from the storage medium to execute the method as claimed in any one of claims 1 to 4.

Images