› Forums › Automatic speech recognition › Dynamic Time Warping (DTW) › Non-English connected speech coarticulation
- This topic has 1 reply, 2 voices, and was last updated 3 years, 9 months ago by Rebekka Puderbaugh.
-
AuthorPosts
-
-
November 6, 2020 at 16:41 #12986
In connected speech, due to coarticulation “it is not possible even to claim the existence of a clear point where one word stops and the next one starts” (Holmes 8.10).
I’m curious whether this is an anglocentric analysis. Learners of English that I know (especially native speakers of Chinese, i.e. an unrelated language, and one with much more restrictive syllable structure rules) often complain to me that we join the words together ‘as if they were one word’ – I wonder if this is because their language doesn’t do so to the same extent. In that case, might the coarticulation effects be much less? (Or maybe we just all think our own language keeps the words distinct, though none does?) -
November 13, 2020 at 12:27 #13048
This appears to be true for all languages and refers mainly to the lack of clear acoustic indicators of word boundaries. That is, if you were to look at a recorded utterance of spontaneous connected speech, you would not be able to identify the beginnings and ends of words based on the acoustics unless you already knew what the words were in the first place. This is true regardless of the language — to my knowledge no language places pauses between words in connected speech, for example.
That being said, languages (and even varieties of a single language) can and do differ in the application of connected speech processes. For example, (some varieties of) English may allow for consonant assimilation across word boundaries while others may not. In Mandarin Chinese, consonants appear to be less affected by such processes, but tones are coarticulated across syllables and words in complex ways (Shen. (1990). Tonal coarticulation in Mandarin https://doi.org/10.1016/S0095-4470(19)30394-8).
For more on cross-language similarities in connected speech processes, you may be interested in this paper (though the sample comparisons are limited to only a few European languages): Barry & Andreeva. (2002). Cross-language similarities and differences in spontaneous speech patterns https://doi.org/10.1017/S0025100301001050
-
-
AuthorPosts
- You must be logged in to reply to this topic.