Abstract
A phonetically-and-prosodically rich sentence set is so important in collecting a read-speech corpus for developing phoneme-based speech recognition. The sentence set is usually searched from a huge text corpus of million sentences using the optimization methods. One of the commonly used optimization methods for this case is a Least-to-Most Greedy (LTMG) algo-rithm. It is effective in minimizing the number of phoneme-units. Unfortunately, it does not distribute their frequencies. In this paper, a new method called Partial LTMG algorithm (PLTMG) is proposed to search an optimum set containing triphones and prosodies those are distributed in a near-uniform fashion. Testing on an Indonesian text corpus of ten million sentences crawled from some websites of newspapers and novels shows that the proposed method is not only capable of minimizing both phoneme-units and prosodies but also effective in distributing their frequencies.
Highlights
Before 2014, an Automatic Speech Recognition (ASR) or Computer Speech Recognition generally has three components, i.e. acoustic model, pronunciation or word lexical model, and language model
The E2EASR does not need both pronunciation and language models commonly used in the previous conventional ASR
The first effort to build an E2EASR system is conducted by some researchers in 2014 using a classification-based approach called Connectionist Temporal Classification (CTC) (Graves 2014)
Summary
Before 2014, an Automatic Speech Recognition (ASR) or Computer Speech Recognition generally has three components, i.e. acoustic model, pronunciation or word lexical model, and language model. The E2EASR does not need both pronunciation and language models commonly used in the previous conventional ASR. The first effort to build an E2EASR system is conducted by some researchers in 2014 using a classification-based approach called Connectionist Temporal Classification (CTC) (Graves 2014) This system consists of a layer of CTC and a Recurrent Neural Networks (RNN), which is abbreviated CTC-RNN. Unlike the CTC-based ASR, this LAS model is capable of learning all ASR components (acoustic models, pronunciation models, and language models) simultaneously. In this paper, a new method called Partial LTM Greedy algorithm (PLTMG) is proposed to search a phonetically-andprosodically rich sentence set with balanced frequencies from an Indonesian text corpus
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: International Journal of Advanced Computer Science and Applications
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.