Robust Word Boundary Detection in Spontaneous Speech using Acoustic and Lexical Clues (bibtex)

by Tsiartas, Andreas, Ghosh, Prasanta Kumar, Georgiou, Panayiotis G. and Narayanan, Shrikanth

Abstract:

We consider the problem of word boundary detection in spontaneous speech utterances. Acoustic features have been well explored in the literature in the context of word boundary detection; however, in spontaneous speech of Switchboard-I corpus, we found that the accuracy of word boundary detec- tion using acoustic features is poor (F-score âˆ¼ 0.63). We pro- pose a new feature - that captures lexical cues in the context of the word boundary detection problem. We show that includ- ing proposed lexical feature along with the usual acoustic fea- tures, the accuracy of the word boundary detection improves considerably (F-score âˆ¼ 0.81). We also demonstrate the ro- bustness of our proposed feature in presence of different noise levels for additive white and pink noise.

View PDF

Reference:

Robust Word Boundary Detection in Spontaneous Speech using Acoustic and Lexical Clues (Tsiartas, Andreas, Ghosh, Prasanta Kumar, Georgiou, Panayiotis G. and Narayanan, Shrikanth), In Proceedings of ICASSP, 2009.

Bibtex Entry:

@inproceedings{tsiartas_robust_2009,
	address = {Taipei, Taiwan},
	title = {Robust {Word} {Boundary} {Detection} in {Spontaneous} {Speech} using {Acoustic} and {Lexical} {Clues}},
	url = {http://ict.usc.edu/pubs/Robust%20Word%20Boundary%20Detection%20in%20Spontaneous%20Speech%20using%20Acoustic%20and%20Lexical%20Clues.pdf},
	abstract = {We consider the problem of word boundary detection in spontaneous speech utterances. Acoustic features have been well explored in the literature in the context of word boundary detection; however, in spontaneous speech of Switchboard-I corpus, we found that the accuracy of word boundary detec- tion using acoustic features is poor (F-score âˆ¼ 0.63). We pro- pose a new feature - that captures lexical cues in the context of the word boundary detection problem. We show that includ- ing proposed lexical feature along with the usual acoustic fea- tures, the accuracy of the word boundary detection improves considerably (F-score âˆ¼ 0.81). We also demonstrate the ro- bustness of our proposed feature in presence of different noise levels for additive white and pink noise.},
	booktitle = {Proceedings of {ICASSP}},
	author = {Tsiartas, Andreas and Ghosh, Prasanta Kumar and Georgiou, Panayiotis G. and Narayanan, Shrikanth},
	month = apr,
	year = {2009},
	keywords = {Graphics}
}