Voice command detection model constructing method, detection method and system, man-machine interaction method and device

Abstract

The invention discloses a voice command detection model constructing method. The method includes determining a typological structure of a voice command detection model; collecting a large amount of man-machine interaction voice data and recognition texts corresponding to the voice data; performing word segmentation on sentence texts of user questions or answers in the recognition text and acquiring word segmentation word strings in the unit of a single-round answers or questions; performing voice command end point evaluation on words in the word segment word string in sequence according to dialogue history information, and marking service types and an information slot or a state slot; by utilizing the recognition text and the evaluation information and mark information of the recognition text, performing training and acquiring parameters of the voice command detection model. The invention also discloses a voice command detection method and system, and a man-machine interaction method and device. The invention can improve voice command point detection accuracy and therefore improves success rate and response speed of machine response in a man-machine interaction process.

Description

technical field [0001] The invention relates to the fields of voice signal processing and human-computer interaction, and in particular to a voice command detection model building method, a voice command detection method and system, and a human-computer interaction method and equipment. Background technique [0002] With the advancement of artificial intelligence technology, human-computer voice interaction has also made great progress, and various voice assistant APPs and human-computer interaction robots are booming. The process by which these voice assistants and bots interact with users is usually as follows: [0003] (1) Wake up by voice or setting operation; [0004] (2) Judge whether the user has finished speaking by the pause duration of the voice data; [0005] (3) After detecting that the user has finished speaking, the recognition result is passed to the semantic understanding engine for semantic understanding; [0006] (4) According to the result of semantic u...

AI Technical Summary

Problems solved by technology

[0008] In view of the problems existing in the above-mentioned existing technology of detecting the end point of voice commands through the pause duration of voice data, the present invention provides a method for constructing a voice command detection model, and a voice command detection method and system based on the voice command detection model. , to avoid the problem of voice commands being truncated due to long pauses inside voice commands, and to prevent interference factors such as noise from affecting the detection of the end point of voice commands

Images