Keyword spotting using a phoneme-sequence index
View Patent ↗In some aspects, a wordspotter is used to locate occurrences in an audio corpus of each of a set of predetermined subword units, which may be phoneme sequences. To locate a query (e.g., a keyword or phrase) in the audio corpus, constituent subword units in the query are indentified and then locations of those subwords are determined based on the locations of those subword units determined earlier by the wordspotter, for example, using a pre-built inverted index that maps subword units to their locations.
1. A method comprising:
accepting a query phrase of a first language;
processing the query phrase using a predefined set of phoneme sequences of a second language to generate a set of phoneme sequences that represent the query phrase;
searching an index that includes locations of instances of a plurality of the phoneme sequence of the generated set to generate a list of candidate locations of the query phrase in one or more media files associated with the index; and
applying a speech recognition algorithm to the one or more media files at the candidate locations of the list;
wherein a second language speech recognition algorithm is applied to the one or more media files at the candidate locations of the list.
2. The method of claim 1 , further comprising:
generating the index based on results of searching a phonetic representation of the one or more media files to locate instances of phoneme sequences of the predefined set of phoneme sequences.
3. A non-transitory computer-readable medium encoded with software comprising instructions for causing a processing system to:
accept a query phrase of a first language;
process the query phrase using a predefined set of phoneme sequences of a second language to generate a set of phoneme sequences that represent the query phrase;
search an index that includes locations of instances of a plurality of the phoneme sequences of the generated set to generate a list of candidate locations of the query phrase in one or more media files associated with the index; and
apply a speech recognition algorithm to the one or more media files at the candidate locations of the list;
wherein a second language speech recognition algorithm is applied to the one or more media files at the candidate locations of the list.
4. The medium of claim 3 , further comprising instructions for causing the processing system to generate the index based on results of searching a phonetic representation of the one or more media files to locate instances of phoneme sequences of the predefined set of phoneme sequences.