Method and device for determining confidence of voice recognition result

Abstract

The invention provides a method and a device for determining the confidence of a voice recognition result, wherein the method comprises the following steps: determining the confidence of each arc in a word graph obtained by decoding, and determining an optimal path in the word graph; determining an arc assembly T competitive to an arc Ai in the word graph for each arc Ai on the optimal path; determining an arc Aj from the arc assembly T competitive to an arc Ai when the confidence of a word expressed by the arc Ai, wherein the arc Aj and the arc Ai represent the same words, or the arc Aj and the arc assembly connected with the arc Aj are formed into a word expressed as same as the arc Ai; determining the confidence of the word represented by the arc Ai according to the confidence of the arc Ai and the confidence of the arc Aj or further according to the confidence of the arc connected with the arc Aj. When the method and the device are used for determining the confidence of the voice recognition result, the component factors of a compound word are considered, so that the confidence can further reflect the real state accurately.

Description

【Technical field】 [0001] The invention relates to the field of speech recognition in computer application technology, in particular to a method and device for determining the confidence level of speech recognition results. 【Background technique】 [0002] In speech recognition, the confidence level is used to indicate the possibility of the recognition result being correct. The larger the value, the higher the possibility of the recognition result being the correct result. It is an important basis for speech recognition. The method for determining the confidence level of the speech recognition result is straightforward. affect the accuracy of speech recognition. [0003] The confidence degree determination of the speech recognition result is mainly obtained by processing the word map (Aattice) generated by decoding. Word graph is a more commonly used expression form of speech recognition results in recent years. It represents multiple decoded candidate results on a directed ...

AI Technical Summary

Problems solved by technology

However, in Chinese, a word can be composed of two other words, so-called compound words, corresponding to this type of words, just as "Chinese people" is composed of the words "China" and "people", the confidence of the existing speech recognition results The degree determination method ignores the constituent factors of compound words, so that the confidence degree of the recognition result cannot reflect the real situation. Since the confidence degree of the recognition result may have an impact on the subsequent adaptive adjustment process of the acoustic model and language model, it will also have an impact on the accuracy of the recognition results

Images