A multi-modal large model-based abstract automatic generation method and system
CN122173643APending Publication Date: 2026-06-09BEIJING GUODIANTONG NETWORK TECH CO LTD +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING GUODIANTONG NETWORK TECH CO LTD
- Filing Date
- 2024-12-06
- Publication Date
- 2026-06-09
AI Technical Summary
Technical Problem
In existing technologies, images in text and image documents may interfere with the large model's understanding of prompt words, leading to inaccurate summary generation. Furthermore, long text content may exceed the model's length limit, affecting the accuracy of summary generation.
Method used
By using a multimodal large model to replace and parse images in text and image documents, full-text data is obtained. Then, text similarity algorithms are used to segment and integrate the text data to generate summary content.
Benefits of technology
It effectively preserves key information in text and image documents, eliminates redundant information, improves the accuracy and efficiency of summary generation, and ensures high-quality summary content.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122173643A_ABST
Abstract
This invention provides an automatic summary generation method and system based on a multimodal large model, comprising: based on a text-image document and its information, calling a multimodal large model to replace and parse the images in the text-image document to obtain full-text data; based on the full-text data, combining a text similarity algorithm to segment and integrate the full-text data to obtain document fragment data; based on the document fragment data, combining a summary prompt word template, calling a question-answering large model to generate summary content of the text-image document. This invention describes the images in the text-image document using a multimodal large model and replaces the images in the document with the image descriptions to retain as much effective information as possible. Then, a text similarity algorithm is used to extract document information fragments containing key content, which can eliminate redundant information in long documents and improve the efficiency and accuracy of subsequent summary generation using the question-answering large model.
Need to check novelty before this filing date? Find Prior Art