Device for signing and extracting original text based on large model generation result comparison

Abstract

The invention discloses a device for signing and extracting an original text based on a large model generation result, and the device is characterized in that a knowledge base management module manages an uploaded file, and analyzes and converts the file into an OFD file; the reasoning question-answering module selects a question-answering model to set a corresponding question-answering prompt word, returns reasoning content and records a file source and an address corresponding to the reasoning content; the original text searching module reads the reasoning result and the original text reference; the diversified endorsement module combines and displays the generated file and the original text, and carries out diversified endorsement and rendering; the combination file generation module synthesizes the signed files to form an OFD file combination file book with the signing effect, and meanwhile, the access of a third-party personnel organization structure is supported, and the files are audited. Based on an artificial intelligence large model and a private knowledge base, a user can initiate services fundamentally, and the user experience is improved. And the knowledge base and the business initiation content are quickly combined by adopting a file combination technology, so that the file organization time of the user is shortened.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and specifically relates to a device for comparing the results generated by a large model with the original text for signature approval and extraction. Background Art

[0002] With the development of the existing technology, the integration degree of existing large models (such as AI Q&A, generative AI) and business systems (such as OA, approval flow) is insufficient, and the large model and the business system cannot achieve business coupling, so that it is impossible to conveniently implement the processing of business such as office work and approval. At the same time, the content of the uploaded files cannot be accurately and conveniently viewed, and the files associated with the Q&A cannot be signed and approved as a whole. Summary of the Invention

[0003] The purpose of the present invention is to provide a device for comparing the results generated by a large model with the original text for signature approval and extraction, so as to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solutions: A device for comparing the results generated by a large model with the original text for signature approval and extraction, including a knowledge base management module, an inference Q&A module, an original text search module, a diversified signature approval module, a combined file generation module, a communication protocol module, and an audit learning module. The knowledge base management module is used for the full life cycle management of the uploaded multimodal files, and generates OFD format files according to the parsing of different file formats; The inference Q&A module is used for selecting the corresponding Q&A model and setting the corresponding Q&A prompt words, and then when returning the inference content, synchronously recording the file source and address corresponding to the inference content; The original text search module is used for referring to and reading the original text according to the inference results displayed on the Q&A interface; The diversified signature approval module is used for merging and displaying the files generated by the above inference Q&A and results with the original text files, and performing diversified signature approval and rendering; The combined file generation module is used for synthesizing multiple files on the terminal for the signed and approved files to form an OFD file combined file with signature approval effects and rendering effects; The communication protocol module is used for carrying the communication protocol for official document transmission and supporting the access of the organizational structure of third-party personnel; The audit learning module is used for auditing different files.

[0005] Preferably, the knowledge base management module performs unified management of the uploaded files, and the uploaded multimodal file formats include wps, doc / docx, xls / xlsx, ppt / pptx, cad, bmp, jpg, tif, gif, png, pdf, html, txt, true, ofd, wmv, rm, mov, mp4, mp3, and avi. The server identifies, classifies, organizes, parses, and converts the uploaded multimodal files, including textualizing the contents of pictures, audio, and video, converting the files into standard OFD format files, and writing the identified OFD metadata content into the OFD format file.

[0006] Preferably, the reasoning question-answering module refers to selecting a corresponding question-answering model, which includes large models deepseek and qianw, setting corresponding question-answering prompt words according to the reasoning content to return the reasoning content, and synchronously recording the file source and address corresponding to the reasoning content.

[0007] Preferably, the original text search module displays the reasoning results of the question on the question-answering interface. By clicking on the reasoning results, the original text can be viewed, and the source content on the original text can be highlighted for reference.

[0008] Preferably, when reading on the question-and-answer interface, choose to form a new file with the content of the question and the generated answer, and attach the searched original content in text form to the location where the question is answered, or include the file in the form of an embedded attachment in the form of OFD, to form a bound file of the processing results or opinions on the leadership's concerns. Clicking on the bound file can jump to the location of the corresponding original file, and the generated bound file and the associated original file are displayed in a combined form.

[0009] Preferably, the diversified approval module supports approval on the generated bound document during approval, and supports image approval, handwritten approval, keyboard input, audio approval, video approval, and rich media approval. It also supports sending the approved content and the original text to the large model for formatting, and content detection and automatic revision of typos and sensitive words. At the same time, in order to allow other users to follow the user's reading ideas, rendering effects can be added to the generated bound document to construct guided document reading.

[0010] Preferably, the communication protocol module directly forwards the signed and approved bound document to the corresponding personnel during the loading process, and chooses to archive it in its own file library. When the file is transmitted, the file is encrypted using a double key method.

[0011] Preferably, for the OFD files sent out for approval, after receiving the files, the recipients need to select different statuses when reading: received, read, understood, executed, and be able to add descriptive languages.

[0012] Preferably, the audit and learning module uses an AI large model to regularly detect the content of the formed files and the content of the approvals for spelling mistakes, sensitivity, stance, and effectiveness; if there are problems with the detected files, the reading is temporarily closed and pushed to the administrator for manual review; for the files generated by inference approvals, a list of official document inference generation is constructed, and the content is reviewed by AI review and manual review by personnel, and the corresponding file content is selected and pushed into the large model or replaced with the file in the knowledge base.

[0013] Compared with the prior art, the beneficial effects of the present invention are: (1) Based on AI inference and answering, combined with diversified approvals, and equipped with a communication protocol, the present invention realizes the business system from query reading to original text verification, to excerpt approval, file combination, sharing instructions (task distribution), and execution feedback, achieving a closed-loop of the business process based on the AI knowledge base.

[0014] (2) Based on the artificial intelligence large model and the user's private knowledge base, the present invention enables users to initiate business with evidence, and uses the file combination technology to quickly combine the knowledge base and the business initiation content, reducing the time for users to organize files and greatly improving work efficiency. Specific embodiments

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Embodiment

[0016] The present invention provides a device for signing and excerpting the original text by comparing the results generated by the large model, including a knowledge base management module, an inference and answering module, an original text search module, a diversified signing and approval module, a combined file generation module, a communication protocol module, and an audit and learning module. Among them, the knowledge base management module performs full-life cycle management on the uploaded multimodal files, such as files with formats of wps, doc / docx, xls / xlsx, ppt / pptx, cad, bmp, jpg, tif, gif, png, pdf, html, txt, true, ofd, wmv, rm, mov, mp4, mp3, avi. The server identifies, classifies, organizes, analyzes, and converts the uploaded multimodal files. Among them, the content in pictures, audio, and video is texturized, and the files are converted into standard OFD format files, and the identified metadata content of OFD is written into the OFD format files; for example, if a doc file is uploaded, the doc file is analyzed and converted into a standard OFD file. Or if the doc file contains pictures, the content in the pictures needs to be texturized so as to better convert it into an OFD file; After the file is converted into an OFD file, the inference and Q&A module selects some Q&A models such as the large model deepseek, qianw, etc. to set corresponding Q&A prompt words. For example, after uploading an official document file, the prompt word is set as: You are a professional intelligent assistant for official document analysis. Please interpret the official document content according to the standard and answer questions. And adjust the similarity between the Q&A questions and the file to be set to 0.65. (The parameters can be adjusted according to requirements). Then, according to the prompt words and the uploaded file, when the inference and Q&A module returns the inference content, it synchronously records the file source and address corresponding to the inference content; When the inference and Q&A module displays the inference content, it displays the inference results of the corresponding questions on the Q&A interface. Through the original text search module, you can click to view the original text next to the inference results and read it with reference to the original text, that is, you can open the original text and highlight the source content for reference reading; when reading on the Q&A interface, select to form a new file with the content of the question and the generated result answer, and append the found original text content in text form at the position of answering the question, or include the file in the form of an OFD embedded attachment in this file to form a bound volume file of the processing results or opinions on matters of concern to the leader. Click on the bound volume file, and you can jump to the position of the corresponding original text file. The generated bound volume file and the associated original text file are displayed in the form of a combined volume. In this way, the generated bound volume file is also an OFD file, or the generated bound volume file and the associated original file can be displayed in the form of an OFD combined volume; The diversified approval module combines the files generated by the original text search module with the original text files for display. It can perform designated approvals, annotations or renderings based on the inferred content. The diversified approval module supports approvals on the generated bound book files. Approvals support operations such as image approvals, handwritten approvals, keyboard input, audio approvals, video approvals, and rich media approvals. It supports file editing with some conventional tools, and supports sending the approved content and the original text to the large model for formatting, as well as content detection and automatic revision of typos and sensitive words. At the same time, in order to allow other users to follow the user's reading ideas, renderings can be added to the generated bound book files and the associated original text files. The system can realize the closed loop of business processes based on AI knowledge base, and the generated bound book file is OFD file. For the OFD file sent out after approval, the recipient needs to select different statuses when reading the file: received, read, understood, executed, and be able to add descriptive language so that the sender can understand the latest status. The combined file generation module synthesizes multiple files at the terminal for the approved files to form an OFD file combined file with approval and rendering effects; The communication protocol module is equipped with a communication protocol for official document transmission, and supports the access of third-party personnel organizational structures. You can select personnel in the organizational structure and directly forward the generated documents after signing and approval to the corresponding personnel, or you can choose to archive them in your own file library. When transferring files, the files are encrypted using a double key method to ensure the security of the file parameters; The audit learning module audits different files. For example, for generated files, the content of the files and the signed and approved content are regularly checked for typos, sensitivity, stance, validity, etc. If a problematic file is detected, the reading is temporarily closed and the file is pushed to the administrator for review. For generated files, a list of official document reasoning is constructed, and the content is audited in a diversified way. You can choose a certain type of file to re-push it, or replace the files in the knowledge base.

[0017] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments are regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention.

Claims

1. An apparatus for signing and extracting by comparing the generated result of a large model with the original text, characterized in that: It includes a knowledge base management module, an inference and Q&A module, an original text search module, a diversified approval module, a combined file generation module, a communication protocol module, and an audit and learning module. The knowledge base management module is used for the full life cycle management of uploaded multimodal files, and generates OFD format files according to different file formats for parsing. The inference and Q&A module is used to select the corresponding Q&A model and set the corresponding Q&A prompt words. When returning the inference content, it synchronously records the file source and address corresponding to the inference content. The original text search module is based on the inference results displayed on the Q&A interface and reads in reference with the original text. The diversified approval module is used to merge and display the files generated by the above inference and Q&A and results with the original text files, and perform diversified approval and rendering. The combined file generation module is used to synthesize multiple files on the terminal for the approved files to form an OFD file combined file with approval effects and rendering effects. The communication protocol module is used to carry the communication protocol for official document transmission and support the access of third-party personnel organizations. The audit and learning module is used to audit different files.

2. The apparatus for signature approval and extraction by comparing the result generated by the large model with the original text according to claim 1, wherein: The knowledge base management module uniformly manages the uploaded files. The formats of the uploaded multimodal files include wps, doc / docx, xls / xlsx, ppt / pptx, cad, bmp, jpg, tif, gif, png, pdf, html, txt, true, ofd, wmv, rm, mov, mp4, mp3, avi. The server identifies, classifies, organizes, analyzes, and converts the uploaded multimodal files. Among them, the content in pictures, audio, and video is texturized, and the files are converted into standard OFD format files, and the identified OFD metadata content is written into the OFD format files.

3. The apparatus for signature approval and extraction by comparing the results generated by the large model with the original text according to claim 1, wherein: The inference and Q&A module refers to selecting the corresponding Q&A model. The Q&A models include large models deepseek and qianw. According to the inference content, the corresponding Q&A prompt words are set to return the inference content, and the file source and address corresponding to the inference content are synchronously recorded.

4. The apparatus for signature approval and extraction by comparing the result generated by the large model with the original text according to claim 1, wherein: The original text search module displays the inference results of the questions on the Q&A interface. Clicking on the inference results can view the original text, and the source content on the original text can be highlighted for reference reading.

5. The apparatus for signature approval and extraction by comparing the result generated by the large model with the original text according to claim 1, wherein: When reading on the Q&A interface, select to form a new file with the content of the question and the generated result answer, and append the found original text content in text form at the position of answering the question, or include the file in the form of an OFD embedded attachment in the file to form a bound volume file of the processing results or opinions on matters of concern to the leadership. Clicking on the bound volume file can jump to the position of the corresponding original text file, and the generated bound volume file and the associated original text file are displayed in the form of a combined volume.

6. The apparatus for signature approval and extraction by comparing the results generated by the large model with the original text according to claim 1, wherein: The diversified approval module supports approval on the generated bound book file during approval, and supports image approval, handwritten approval, keyboard input, audio approval, video approval, and rich media approval. It also supports sending the approved content and the original text to the large model for formatting, and content detection and automatic revision of typos and sensitive words. At the same time, in order to allow other users to follow the user's reading ideas, rendering effects can be added to the generated bound book file and the associated original text file to build a guided file reading.

7. The apparatus for signature approval and extraction by comparing the result generated by the large model with the original text according to claim 1, wherein: The communication protocol module directly forwards the signed and bound document to the corresponding personnel during the loading process, and chooses to archive it into its own file library. When the file is transmitted, the file is encrypted using a double key method.

8. The apparatus for signature approval and extraction by comparing the results generated by the large model with the original text according to claim 5, wherein: The generated bound document is an OFD document. For the OFD document that is signed and sent out, the recipient needs to select different statuses when reading the document after receiving it: received, read, understood, executed, and can add descriptive language so that the sender can understand the latest status.

9. The apparatus for signature approval and extraction by comparing the results generated by the large model with the original text according to claim 1, wherein: The audit learning module uses the AI ​​big model to regularly check the content of the files and the approved content for typos, sensitivity, stance, and validity. If the detected file has problems, the reading is temporarily closed and pushed to the administrator for manual review. For the files generated by reasoning and approval, a list of official document reasoning is constructed, and the content is reviewed by AI review and manual review by personnel, and the corresponding file content is selected and pushed to the big model, or the file in the knowledge base is replaced.