Investigation into Transferability of Duration of Emphasised Words from Original Expression to Spoken Translation 梁

To achieve an expressive, personalised speech-to-speech translator, paralinguistic information need be transferred from input natural speech into output synthesised speech, in particular, a user’s focuses carried by emphasised spoken words in input speech. In order to examine if the duration of emphasised words in a source language is usable for predicting that of their emphasised spoken translations, this paper presents results of analysis of sentence-level and word-level duration extracted from parallel English and German speech utterances containing emphasis. Firstly, it was confirmed to be reasonable to consider emphasised words did not affect duration of their nearby neutral parts. Secondly, the realisation of emphasis in terms of word duration was speaker-dependent to an extent and language-independent for most speakers. Thirdly, though the duration of an emphasised word in a source language by itself was not sufficient for a definite prediction of the duration of its spoken translation in a target language, collectively, duration of emphasised words in the source language could still portray a speaker’s own style of emphasising words in the target language.