An improved model for recognizing disfluencies in conversational speech

Mark Johnson; Eugene Charniak; Matthew Lease

An improved model for recognizing disfluencies in conversational speech

Mark Johnson, Eugene Charniak, Matthew Lease

Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding contribution › peer-review

Abstract

This paper presents a novel metadata extraction (MDE) system for automatically detecting edited words, fillers, and self-interruption points in conversational speech. Our edit word detection sub-system combines a Tree Adjoining Grammar (TAG) noisy channel model, a statistical syntactic language model, and a MaxEnt reranker. Hand-built, deterministic rules are used to detect fillers. Self-interruption points are explicitly determined by detected fillers and edited words. We have evaluated our system for these three tasks on two types of input: manually annotated words and automatically recognized speech-to-text tokens. In all six cases, our system has improved the state-of-the-art, as measured in a recent blind evaluation.

Original language	English
Title of host publication	Rich Transcription 2004 Fall workshop (RT-04F)
Place of Publication	Palisades, NY
Number of pages	7
Publication status	Published - 2004
Externally published	Yes
Event	Rich Transcription 2004 Fall workshop - Palisades, United States Duration: 7 Nov 2004 → 10 Nov 2004

Conference

Conference	Rich Transcription 2004 Fall workshop
Abbreviated title	RT-04F
Country/Territory	United States
City	Palisades
Period	7/11/04 → 10/11/04

Cite this

@inproceedings{b99982d3852a47ac822a2401ae49b4a5,

title = "An improved model for recognizing disfluencies in conversational speech",

abstract = "This paper presents a novel metadata extraction (MDE) system for automatically detecting edited words, fillers, and self-interruption points in conversational speech. Our edit word detection sub-system combines a Tree Adjoining Grammar (TAG) noisy channel model, a statistical syntactic language model, and a MaxEnt reranker. Hand-built, deterministic rules are used to detect fillers. Self-interruption points are explicitly determined by detected fillers and edited words. We have evaluated our system for these three tasks on two types of input: manually annotated words and automatically recognized speech-to-text tokens. In all six cases, our system has improved the state-of-the-art, as measured in a recent blind evaluation.",

author = "Mark Johnson and Eugene Charniak and Matthew Lease",

year = "2004",

language = "English",

booktitle = "Rich Transcription 2004 Fall workshop (RT-04F)",

note = "Rich Transcription 2004 Fall workshop, RT-04F ; Conference date: 07-11-2004 Through 10-11-2004",

}

TY - GEN

T1 - An improved model for recognizing disfluencies in conversational speech

AU - Johnson, Mark

AU - Charniak, Eugene

AU - Lease, Matthew

PY - 2004

Y1 - 2004

N2 - This paper presents a novel metadata extraction (MDE) system for automatically detecting edited words, fillers, and self-interruption points in conversational speech. Our edit word detection sub-system combines a Tree Adjoining Grammar (TAG) noisy channel model, a statistical syntactic language model, and a MaxEnt reranker. Hand-built, deterministic rules are used to detect fillers. Self-interruption points are explicitly determined by detected fillers and edited words. We have evaluated our system for these three tasks on two types of input: manually annotated words and automatically recognized speech-to-text tokens. In all six cases, our system has improved the state-of-the-art, as measured in a recent blind evaluation.

AB - This paper presents a novel metadata extraction (MDE) system for automatically detecting edited words, fillers, and self-interruption points in conversational speech. Our edit word detection sub-system combines a Tree Adjoining Grammar (TAG) noisy channel model, a statistical syntactic language model, and a MaxEnt reranker. Hand-built, deterministic rules are used to detect fillers. Self-interruption points are explicitly determined by detected fillers and edited words. We have evaluated our system for these three tasks on two types of input: manually annotated words and automatically recognized speech-to-text tokens. In all six cases, our system has improved the state-of-the-art, as measured in a recent blind evaluation.

M3 - Conference proceeding contribution

BT - Rich Transcription 2004 Fall workshop (RT-04F)

CY - Palisades, NY

T2 - Rich Transcription 2004 Fall workshop

Y2 - 7 November 2004 through 10 November 2004

ER -

An improved model for recognizing disfluencies in conversational speech

Abstract

Conference

Fingerprint

Cite this