A Speech Recognition Infinite Level Multi-Turn Dialogue Method

Abstract

The invention discloses a voice recognition infinite-level multi-round dialogue method. A function card model file is formulated. The card types of the function card model file include main cards and alternative cards. After the voice input and recognition results are obtained, the recognized The semantics are matched with the corresponding semantics of the cards one by one. After the cards are matched, the corresponding card instructions are executed and the matching candidate cards are reloaded into the pool to be matched according to the corresponding next-level dialogue ID. The first round of dialogue will The main card stored in the corresponding pool to be matched is used as the matching object, and then each round of dialogue takes the reloaded candidate card after the previous round of dialogue matching as the matching object. When the matched card has no next-level dialogue ID or undefined End the dialog on exception. No matter how many levels of dialogues the present invention needs to match, the number of matching targets is only related to the upper level of dialogues, and the increase of the number of associated dialogues has no effect on the matching efficiency, which ensures the matching efficiency and reliability of multi-level infinite rounds of dialogues.

Description

technical field [0001] The invention relates to a speech recognition infinitely multi-round dialogue method. Background technique [0002] At present, the application range of voice recognition dialogue control equipment is getting wider and wider, including navigation equipment, which also begins to use this method for control and automatic answering. In a multi-round dialogue, a dialogue is not only related to the requirements of the dialogue itself, but also needs to be combined with the content of the previous dialogue to determine the correct feedback and response corresponding to the dialogue. Compared with a single dialogue, this is called a second-level dialogue. The more dialogue levels, the higher the number of levels. However, the current speech recognition multi-level dialogues on the market are at most three-level dialogues, that is, dialogues that can be continuously related to the previous dialogue content can only contain three levels. This is because the exi...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a method for speech recognition with infinite levels of multi-round dialogues, to solve the limitations of associated multi-level dialogues in the prior art due to the large number of targets and associated content involved in matching, low matching efficiency, high error rate, etc. dialogue progression

Images