Characteristic and split data feature processing method and related equipment

By processing land acquisition and demolition data using a pre-defined multimodal model, multi-dimensional variable-dimensional semantic vectors are generated, solving the problem of insufficient semantic expression in multimodal scenarios by traditional models and achieving more efficient data processing.

CN121837428APending Publication Date: 2026-04-10GUANGZHOU ZHENGYUN TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-25
Publication Date
2026-04-10

AI Technical Summary

Technical Problem

Traditional vector models lack the semantic vector representation capability in multimodal land acquisition and demolition data scenarios, making it difficult to meet the accuracy and efficiency requirements of land acquisition and demolition business data processing.

Method used

A pre-defined multimodal model is adopted, including a pre-defined text encoder, image encoder, feature fusion layer, hybrid expert model and MRL representation learning layer, to generate multi-dimensional variable-dimensional semantic vectors through feature extraction, fusion and nested representation processing.

Benefits of technology

It improves the semantic vector representation capability of land acquisition and demolition data, enhances the accuracy and efficiency of data processing, and reduces model performance loss.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121837428A_ABST
    Figure CN121837428A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a characterization and division data feature processing method and related equipment, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring to-be-processed eigen-split data, and inputting the to-be-processed eigen-split data into a preset multi-modal model for feature processing to obtain a target variable nested dimension vector; wherein feature extraction is carried out on the feature division text data through a preset text encoder to obtain text feature data, feature extraction is carried out on the feature division image data through a preset image encoder to obtain image feature data, and then feature fusion is carried out through a feature fusion layer to obtain a preset feature sequence; and performing feature learning on the preset feature sequence through a preset hybrid expert model to generate a full-dimension feature vector, and performing nested representation processing through an MRL representation learning layer to generate a target variable nested dimension vector. According to the embodiment of the invention, the generation of the multi-dimensional variable-dimension semantic vector of the eigen-split data can be realized, the semantic vector expression capability of the eigen-split data is improved, and the accuracy and efficiency of eigen-split data processing are improved.
Need to check novelty before this filing date? Find Prior Art