Speech recognition method and apparatus
AssigneeAlibaba Group Holding Limited
InventorsXiaohui Li, Hongyan Li
Life of this patent
Estimating the term…
Abstract
A speech recognition method comprises: generating, based on a preset speech knowledge source, a search space comprising preset client information and for decoding a speech signal; extracting a characteristic vector sequence of a to-be-recognized speech signal; calculating a probability at which the characteristic vector corresponds to each basic unit of the search space; and executing a decoding operation in the search space by using the probability as an input to obtain a word sequence corresponding to the characteristic vector sequence.
Claims20
A method, comprising:
- acquiring, through decoding, a word sequence corresponding to a speech signal;
- and
- verifying the accuracy of the word sequence and generating a speech recognition result by performing operations comprising:
- selecting from the word sequence a first word that corresponds to client information;
- searching for the first word in the client information;
- if the first word is found in the client information, using the word sequence as the speech recognition result;
- and
- if the first word is not found in the client information, correcting the word sequence by pinyin-based fuzzy matching, and using the corrected word sequence as the speech recognition result.
One or more non-transitory computer-readable mediums storing instructions executable by one or more processors, wherein execution of the instructions cause the one or more processors to perform operations comprising:
- acquiring, through decoding, a word sequence corresponding to a speech signal;
- and
- verifying the accuracy of the word sequence and generating a speech recognition result by performing operations comprising:
- selecting from the word sequence a first word that corresponds to client information;
- searching for the first word in the client information;
- if the first word is found in the client information, using the word sequence as the speech recognition result;
- and
- if the first word is not found in the client information, correcting the word sequence by pinyin-based fuzzy matching, and using the corrected word sequence as the speech recognition result.
A system, comprising:
- one or more processors;
- one or more non-transitory computer-readable mediums storing instructions executable by the one or more processors, wherein execution of the instructions cause the one or more processors to perform operations comprising:
- acquiring, through decoding, a word sequence corresponding to a speech signal;
- and
- verifying the accuracy of the word sequence and generating a speech recognition result by performing operations comprising:
- selecting from the word sequence a first word that corresponds to client information;
- searching for the first word in the client information;
- if the first word is found in the client information, using the word sequence as the speech recognition result;
- and
- if the first word is not found in the client information, correcting the word sequence by pinyin-based fuzzy matching, and using the corrected word sequence as the speech recognition result.
Dependencies are inferred from the claim text.