Abstract

Linguistic and perceptual arguments suggest that, in speech recognition systems, syntactic hypotheses should be formed before phonemic segments are identified. Prosodic features can provide some cues to constituent structure. In a variety of texts and excerpts from conversations, spoken by several talkers, a decrease in voice fundamental frequency (F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> ) usually occurred at the end of each major syntactic constituent, and an increase in F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> occurred near the beginning of the following constituent. A computer program based on this regularity correctly detected over 80 percent of all syntactically predicted boundaries. Some boundaries between minor constituents were also detected by the fall-rise patterns in F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> . False boundary detections resulted from F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> variations at boundaries between vowels and consonants, but most such false alarms could be eliminated by setting a minimum percent variation in F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> for a boundary detection. Sentence boundaries were accompanied by large F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf> increases and substantial pauses. The categories of constituents affect boundary detection results, with noun phrase-verbal sequences showing particularly infrequent detection. Prosodic cues to stress patterns and stress-to-syntax rules may be used to detect other aspects of syntactic structure. Syntactic structure hypotheses might then be used to guide phonetic recognition procedures within constituents.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

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.