Automatic selection of phonetically distributed sentence sets for speaker adaptation with application to large vocabulary Mandarin speech recognition

Automatic recognition of oral vowels in tone language: Experiments with fuzzy logic and neural network models

“…Some works have also been reported on Asian tone languages such as Mandarin [6,14,16,19,23,31,32] Cantonese [15,21,22,25,63] and Thai [39,59]. Ray and Ghoshal [26] have also reported the development an ASR for vowels of three Indian languages, namely: Telugu, Assamese and Bengali and Salor et al [60] have reported works on Turkish.…”

Designing text corpus using phone-error distribution for acoustic modeling

“…There have been also several studies for the latter approach to provide a good text corpus to collect read speech data [6], [7], [8], [9]. Shen et al [7] proposed to design a phonetically balanced sentence set.…”

“…Shen et al [7] proposed to design a phonetically balanced sentence set. While this approach is useful to avoid the data sparseness problem, it does not directly increase the recognition performance.…”

A Combined Task Analysis Method for Data Selection in Mandarin Isolated Word Recognition System

“…In [1], adaptation script is generated with two stages including 1) coverage unit distribution derived by task analysis; 2) script generation based on task dependent coverage unit distribution. In [2], all context-independent (CI) INITIAL/FINAL in the task are used as the coverage units and the distribution of the coverage unit occurrence in the task then obtained. The adaptation script is generated by covering all coverage units with minimum number of the sentences firstly and approximating the distribution of the coverage units by the adaptation script secondly.…”