Large-Scale clustering of acoustic segments for sub-word acoustic modelling

Simple item page
dc.contributor.advisorNiesler, T. R.en_ZA
dc.contributor.authorLerato, Leratoen_ZA
dc.contributor.otherStellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.en_ZA
dc.date.accessioned2019-02-01T07:48:22Z
dc.date.accessioned2019-04-17T08:11:43Z
dc.date.available2019-02-01T07:48:22Z
dc.date.available2019-04-17T08:11:43Z
dc.date.issued2019-04
dc.descriptionThesis (PhD)--Stellenbosch University, 2019.en_ZA
dc.description.abstractENGLISH ABSTRACT: A pronunciation dictionary is one of the key building blocks in automatic speech recognition (ASR) systems. However, pronunciation dictionaries used in state-of-the-art ASR systems are hand-crafted by linguists. This process requires expertise, time and funding and as a consequence is not realised for many under-resourced languages. To address this, we develop a new unsupervised agglomerative hierarchical clustering (AHC) algorithm that can be used to discover sub-word units that can in turn be used for the automatic induction of a pronunciation dictionary. The new algorithm, named multi-stage agglomerative hierarchical clustering (MAHC), addresses the O(N2) memory and computation complexity observed when classical AHC is applied to large datasets. MAHC splits the data into independent subsets and applies AHC to each. The resultant clusters are merged, re-divided into subsets, and passed to a following iteration. Results show that MAHC can match and even surpass the performance of classical AHC. Furthermore, MAHC can automatically determine the optimal number of clusters which is a feature not offered by most other approaches. A further refinement of MAHC, termed MAHC with memory size management (MAHC+M), addresses the case where some subsets may exhibit excessive growth during iterative clustering. MAHC+M is able to adhere to maximum memory constraints, which improves efficiency and is practically useful when using parallel computing resources. The input to MAHC is a matrix of pairwise distances computed with dynamic time warping (DTW). A modified form of DTW, named feature trajectory DTW (FTDTW), is introduced and shown to generally lead to better performance for both MAHC and MAHC+M. It is shown that clusters obtained using the MAHC algorithm can be used as sub-word units (SWUs) for acoustic modelling. Pronunciations in terms of these SWUs were obtained by alignment with the orthography. Speech recognition experiments show that dictionaries induced using clusters obtained by FTDTW-based MAHC+M consistently outperform those obtained using DTW-based MAHC.en_ZA
dc.format.extent125 pagesen_ZA
dc.identifier.urihttp://hdl.handle.net/10019.1/105757
dc.language.isoen_ZAen_ZA
dc.publisherStellenbosch : Stellenbosch Universityen_ZA
dc.rights.holderStellenbosch Universityen_ZA
dc.subjectLarge-Scale Clustering; Acoustic Segments; Sub-word; Acoustic Modellingen_ZA
dc.subjectAutomatic speech recognitionen_ZA
dc.subjectAgglomerationsen_ZA
dc.subjectAcoustical engineeringen_ZA
dc.subjectUCTDen_ZA
dc.titleLarge-Scale clustering of acoustic segments for sub-word acoustic modellingen_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
lerato_scale_2019.pdf
Size:
5.38 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Plain Text
Description:
Download
Collections
Doctoral Degrees (Electrical and Electronic Engineering)