Report automatic generation method and device based on multi-modal large model and electronic equipment

Abstract

This application provides a method, apparatus, and electronic device for automatically generating reports based on a multimodal large model, comprising: acquiring historical complaint handling data; constructing a multimodal knowledge base and multiple training sets by performing data cleaning and normalization, multi-dimensional classification and labeling, and multi-form data augmentation on the historical complaint handling data; training a multimodal large model based on the multimodal knowledge base and multiple training sets; automatically retrieving historical complaint handling data matching the new complaint data from the multimodal knowledge base through similarity analysis and comparison; extracting key elements from the retrieved historical complaint handling data and inputting the key elements into the multimodal large model to obtain new complaint handling data output by the multimodal large model; and automatically generating a multi-dimensional analysis report of the new complaint handling data based on the new complaint handling data. This method is used to solve the problems of inaccurate root cause analysis and delayed rectification measures in existing technologies.

Images