Trang Tran
University of Washington
20 Papers
61 Citations
Trang Tran is an academic researcher from University of Washington. The author has contributed to research in topics: Prosody & Parsing. The author has an hindex of 8, co-authored 17 publications. Previous affiliations of Trang Tran include Bucknell University & University of Southern California.
Chat about Author
Papers
Parsing Speech: a Neural Approach to Integrating Lexical and Acoustic-Prosodic Information
Trang Tran,Shubham Toshniwal,Mohit Bansal,Kevin Gimpel,Karen Livescu,Mari Ostendorf +5 more
- 01 Jun 2018
TL;DR: A model that integrates transcribed text and acoustic-prosodic features using a convolutional neural network over energy and pitch trajectories coupled with an attention-based recurrent neural network that accepts text and prosodic features is introduced.
Disfluencies and Human Speech Transcription Errors.
Vicky Zayats,Trang Tran,Richard Wright,Courtney Mansfield,Mari Ostendorf +4 more
- 15 Sep 2019
TL;DR: This article explored contexts associated with errors in transcrip-tion of spontaneous speech, shedding light on human perceptionof disfluencies and other conversational speech phenomena, and provided disfluency annotations for careful speech transcripts.
32
•Posted Content
Parsing Speech: A Neural Approach to Integrating Lexical and Acoustic-Prosodic Information
TL;DR: The authors used a convolutional neural network over energy and pitch trajectories coupled with an attention-based recurrent neural network that accepted text and prosodic features for automatically parsing spoken utterances, and found that different types of acoustic-prosodic features are individually helpful and together give statistically significant improvements in parse and disfluency detection F1 scores over a strong text-only baseline.
25
On the Role of Style in Parsing Speech with Neural Models.
Trang Tran,Jiahong Yuan,Yang Liu,Mari Ostendorf +3 more
- 15 Sep 2019
TL;DR: It is shown that neural approaches facilitate using written text to improve parsing of spontaneous speech, and that prosody further improves over this state-of-the-art result, with spontaneous speech more generally useful for training parsers.
•Posted Content
Disfluencies and Human Speech Transcription Errors
TL;DR: This paper explored contexts associated with errors in transcrip-tion of spontaneous speech, shedding light on human perceptionof disfluencies and other conversational speech phenomena, and provided disfluency annotations for careful speech transcripts.