一种文本召回方法、装置、存储介质及设备
By filtering and grouping texts, and using a variable classification relevance recognition model to calculate the relevance between text and labels, the problem of low accuracy in the relevance representation of text and three-level labels in existing technologies is solved, and more efficient relevance judgment and calculation are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- MICRO DREAM TECHTRONIC NETWORK TECH CHINACO
- Filing Date
- 2022-12-23
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, the accuracy of relevance representation between text and third-level tags in Weibo three-level tag recall methods is low, and existing rule strategies are difficult to effectively represent the relevance between text and third-level tags.
By determining the relevance score between text and label, texts with a first score greater than the threshold are selected and grouped. A variable classification relevance recognition model is used to calculate the relevance between text and label within the label group, and texts with a second score greater than the threshold are output as target texts.
It improves the accuracy and computational efficiency of text-tag relevance judgment, reduces reliance on tag matching rules, lowers maintenance costs, and can calculate the relevance of multiple texts and tags simultaneously.
Smart Images
Figure CN116108173B_ABST