Abstract | ||
---|---|---|
Negative transfer in training of acoustic models for automatic speech recognition has been reported in several contexts such as domain change or speaker characteristics. This paper proposes a novel technique to overcome negative transfer by efficient selection of speech data for acoustic model training. Here data is chosen on relevance for a specific target. A sub modular function based on likelihood ratios is used to determine how acoustically similar each training utterance is to a target test set. The approach is evaluated on a wide domain data set, covering speech from radio and TV broadcasts, telephone conversations, meetings, lectures and read speech. Experiments demonstrate that the proposed technique both finds relevant data and limits negative transfer. Results on a 6 hour test set show a relative improvement of 4% with data selection over using all data in PLP based models, and 2% with DNN features. |
Year | Venue | Keywords |
---|---|---|
2015 | 16TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION (INTERSPEECH 2015), VOLS 1-5 | data selection, transfer learning, negative transfer, speech recognition |
Field | DocType | Volume |
Negative transfer,Voice activity detection,Computer science,Transfer of learning,Utterance,Submodular set function,Speech recognition,Multi domain,Artificial intelligence,Machine learning,Acoustic model,Test set | Journal | abs/1509.02409 |
ISSN | Citations | PageRank |
16th Interspeech.Proc. (2015) 2897-2901 | 5 | 0.39 |
References | Authors | |
18 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Mortaza Doulaty | 1 | 33 | 5.35 |
Oscar Saz | 2 | 142 | 16.30 |
Thomas Hain | 3 | 171 | 28.29 |