Abstract | ||
---|---|---|
Nonverbal vocalizations are one of the characteristics of spontaneous speech distinguishing it from written text. These phenomena are sometimes regarded as a problem in language and acoustic modeling. However, vocalizations such as filled pauses enhance language models at the local level and serve some additional functions (marking linguistic boundaries, signaling hesitation). In this paper we investigate a wider range of nonverbals and investigate their potential for language modeling of conversational speech, and compare different modeling approaches. We find that all nonverbal sounds, with the exception of breath, have little effect on the overall results. Due to its specific nature, as well as its frequency in the data, modeling of breath as a regular language model event leads to a substantial improvement in both perplexity and speech recognition accuracy. |
Year | DOI | Venue |
---|---|---|
2012 | 10.1007/978-3-642-32790-2_59 | Lecture Notes in Computer Science |
Keywords | Field | DocType |
regular language,data model,language model,speech recognition | Perplexity,Computer science,Word error rate,Cued speech,Speech recognition,Nonverbal communication,Natural language processing,Artificial intelligence,Regular language,Language model | Conference |
Volume | ISSN | Citations |
7499 | 0302-9743 | 1 |
PageRank | References | Authors |
0.48 | 10 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Dmytro Prylipko | 1 | 66 | 4.65 |
Bogdan Vlasenko | 2 | 235 | 12.72 |
Andreas Stolcke | 3 | 6690 | 712.46 |
Andreas Wendemuth | 4 | 451 | 41.74 |