Industrial domain event joint extraction method, device and equipment based on feature fusion

By combining feature fusion and neural networks, the problems of insufficient error propagation and dependency capture in existing technologies are solved, achieving high accuracy in event extraction in the industrial field.

CN117332073BActive Publication Date: 2026-06-12CHINA POWER IND INTERNET CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
CHINA POWER IND INTERNET CO LTD
Filing Date
2023-10-09
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

Existing event extraction techniques suffer from error propagation problems, leading to decreased accuracy and an inability to effectively capture the interdependencies among event participants.

⚗Method used

We adopt a joint event extraction method for industrial fields based on feature fusion. This method obtains and fuses multiple features of text data (hidden layer features, magnetic features, dependency syntax features, word vector features, and character-level convolutional features), uses attention mechanism and recurrent neural network for semantic feature extraction, and combines gating mechanism and feedforward neural network for event extraction. We construct an encoder-decoder framework to capture the dependencies between events.

🎯Benefits of technology

It improves the accuracy of event extraction, reduces error propagation, enhances the capture of interdependencies among event participants, and improves the overall accuracy of event extraction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117332073B_ABST
    Figure CN117332073B_ABST
Patent Text Reader

Abstract

The application relates to an industrial field event joint extraction method, device and equipment based on feature fusion. Different features of each word in text data to be processed are extracted, and the features are fused to obtain a fusion feature sequence corresponding to the text data. A hidden layer semantic feature sequence is obtained by extracting semantic features according to the fusion feature sequence. Then, a learning neural network is used to decode the hidden layer semantic feature sequence. Trigger words in an event frame are extracted first, then event parameters are determined according to the trigger words, the event category of the event frame is determined according to the event parameters, and finally the role label is determined according to the event category, so that event extraction of the text data is realized. The method can improve the accuracy of event extraction.
Need to check novelty before this filing date? Find Prior Art