Paper: Detecting Structural Metadata With Decision Trees And Transformation-Based Learning

ACL ID N04-1018
Title Detecting Structural Metadata With Decision Trees And Transformation-Based Learning
Venue Human Language Technologies
Session Main Conference
Year 2004
Authors

The regular occurrence of dis uencies is a distinguishing characteristic of spontaneous speech. Detecting and removing such dis u- encies can substantially improve the usefulness of spontaneous speech transcripts. This pa- per presents a system that detects various types of dis uencies and other structural information with cues obtained from lexical and prosodic information sources. Speci cally, combina- tions of decision trees and language models are used to predict sentence ends and interruption points and, given these events, transformation- based learning is used to detect edit dis uen- cies and conversational llers. Results are re- ported on human and automatic transcripts of conversational telephone speech.