Deriving disyllabic word variants from a Chinese conversational speech corpus

Abstract

Motivated by the quasi-categorical reduced forms of disyllabic words produced in Chinese conversational speech, a frequency-based selection procedure of typical pronunciation by disyllabic word type and reduction degree is proposed in this paper. This variant-selection algorithm utilizes techniques of free phone recognition and phonetic similarity score calculation to account for Chinese syllable structure. Four reduction types are suggested by considering the presence of a within-word syllable boundary: Citation form-like reduction, marginal segment deletion, nuclei merger, and syllable merger. The results show that the most frequent reduction types for disyllabic words in Chinese conversation are citation form-like reduction and syllable merger. In particular, high-frequency disyllabic words preferentially take the extreme syllable-merger form. As shown in the analysis, segmental reduction in Chinese disyllabic words is morphology-dependent. It is also related to the prosodic position at which a disyllabic word is produced as well as the temporal quality of the word. Finally, in the automatic speech recognition experiments, the performance was improved by adding a small number of variants selected by the algorithm to the pronunciation dictionary of the system.

Proceedings of DiSS 2019

“…However, in spite of similar duration patterns, NH children clearly differentiate themselves from CI and HA children in the analysis of word-level reduction types. The general pattern is similar to that reported in Liu et al (2016) for adults' speech, where CAN and SYM are preferred than MSD and NUM. NH children apparently produced more CAN and less SYM than both of the CI and HA children ( Figure 5).…”

“…In order to examine word category from a more general perspective, we transformed all CKIP POS tags to Universal POS categories (Universal Dependencies webpage). For word-level reduction, we adopted the derivation approach proposed by Liu et al (2016) by comparing surface phone sequences generated from free phone recognition results with the phoneme sequences directly derived from the standard phonological forms of the disyllabic words. Four reduction degrees are differentiated from minor to severe: CAN < MSD < NUM < SYM.…”

[Retracted] Russian Speech Conversion Algorithm Based on a Parallel Corpus and Machine Translation

“…The volume of translation is enormous, far exceeding the capabilities of pure manual labor, and terminology and translation must be consistent. It not only raises new questions about the feasibility and in-depth research of machine translation and computer-assisted translation but also creates a significant market for the latter [6]. In this sense, the industrialization of translation is inextricably linked to information technology, and the advancement of information technology has aided in the formation of a large translation market, which has become the reason for the existence of large translation service enterprises [7].…”