A method for estimating natural language understanding confidence

By combining speech recognition and natural language understanding, and fusing sentence components and speech segment confidence, the problem of inaccurate confidence estimation in existing technologies is solved, thus improving the accuracy of human-computer interaction.

CN119207390BActive Publication Date: 2026-06-23PACHIRA TIMES (ZHUHAI HENGQIN) INFORMATION TECH CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
PACHIRA TIMES (ZHUHAI HENGQIN) INFORMATION TECH CO LTD
Filing Date
2024-09-19
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing natural language understanding technologies fail to effectively combine information such as confidence level, tone, volume, and emotion of pronunciation segments during speech recognition, resulting in inaccurate confidence estimation and affecting the accuracy of human-computer interaction.

Method used

Speech recognition is performed by acquiring speech information, and the confidence of each recognized segment is calculated by combining sentence components such as prefixes, suffixes, modal particles and action intentions. This confidence is then fused with the natural language understanding decoding path, giving higher weight to the key parts of the sentence to obtain the final confidence score.

Benefits of technology

It improves the accuracy of confidence estimation in natural language understanding and enhances the success rate of human-computer interaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119207390B_ABST
    Figure CN119207390B_ABST
Patent Text Reader

Abstract

The application discloses a natural language understanding confidence estimation method, which mainly comprises the following steps: acquiring voice information of a user, performing voice recognition on the voice information to obtain recognized text information; performing natural language understanding decoding on the recognized text information, and giving a decoding path confidence of each decoding path for the recognized text information in combination with a sentence component; acquiring a confidence corresponding to each recognized segment in a voice recognition process; and fusing the confidence corresponding to each recognized segment and the decoding path confidence on each decoding path of the natural language understanding to obtain a final confidence of each decoding path. The application effectively uses the confidence information of each recognized result segment in the voice recognition process, simultaneously considers that a key part in a sentence has a greater influence on the confidence, can more accurately grasp the intention of the voice, improves the accuracy of the semantics, and further improves the success rate of human-computer interaction.
Need to check novelty before this filing date? Find Prior Art