Multi-accent acoustic modelling of South African English

Simple item page
dc.contributor.authorKamper H.
dc.contributor.authorMuamba Mukanya F.J.
dc.contributor.authorNiesler T.
dc.date.accessioned2012-05-17T08:58:55Z
dc.date.available2012-05-17T08:58:55Z
dc.date.issued2012
dc.description.abstractAlthough English is spoken throughout South Africa it is most often used as a second or third language, resulting in several prevalent accents within the same population. When dealing with multiple accents in this under-resourced environment, automatic speech recognition (ASR) is complicated by the need to compile multiple, accent-specific speech corpora. We investigate how best to combine speech data from five South African accents of English in order to improve overall speech recognition performance. Three acoustic modelling approaches are considered: separate accent-specific models, accent-independent models obtained by pooling training data across accents, and multi-accent models. The latter approach extends the decision-tree clustering process normally used to construct tied-state hidden Markov models (HMMs) by allowing questions relating to accent. We find that multi-accent modelling outperforms accent-specific and accent-independent modelling in both phone and word recognition experiments, and that these improvements are statistically significant. Furthermore, we find that the relative merits of the accent-independent and accent-specific approaches depend on the particular accents involved. Multi-accent modelling therefore offers a mechanism by which speech recognition performance can be optimised automatically, and for hard decisions regarding which data to pool and which to separate to be avoided. © 2012 Elsevier B.V. All rights reserved.
dc.identifier.citationSpeech Communication
dc.identifier.citation54
dc.identifier.citation6
dc.identifier.citation801
dc.identifier.citation813
dc.identifier.issn1676393
dc.identifier.other10.1016/j.specom.2012.01.008
dc.identifier.urihttp://hdl.handle.net/10019.1/21010
dc.subjectAcoustic modelling
dc.subjectAutomatic speech recognition
dc.subjectClustering process
dc.subjectHard decisions
dc.subjectHidden markov models (HMMs)
dc.subjectSouth Africa
dc.subjectSouth African English accents
dc.subjectSpeech corpora
dc.subjectSpeech data
dc.subjectSpeech recognition performance
dc.subjectTraining data
dc.subjectWord recognition
dc.subjectClustering algorithms
dc.subjectHidden Markov models
dc.subjectSeparation
dc.subjectSpeech recognition
dc.titleMulti-accent acoustic modelling of South African English
dc.typeArticle
Files
Collections
Stellenbosch University - Scopus Publications