Abstract

Corpus-based grammar induction relies on using many hand-parsed sentences as training examples. However, the construction of a training corpus with detailed syntactic analysis for every sentence is a labor-intensive task. We propose to use sample selection methods to minimize the amount of annotation needed in the training data, thereby reducing the workload of the human annotators. This paper shows that the amount of annotated train- ing data can be reduced by 36% without degrading the quality of the induced grammars.

Citations

6514	Elements of Information Theory - Cover, Thomas - 1991
1654	B.: Building a large annotated corpus of english: The Penn treebank - Marcus, Marcinkiewicz, et al. - 1993
427	Three Generative, Lexicalised Models for Statistical Parsing - Collins - 1997
365	A sequential algorithm for training text classifiers - Lewis, Gale - 1994
334	Improving generalization with active learning - Cohn, Atlas, et al. - 1994
324	Statistical Parsing with a Context-Free Grammar and Word Statistics - Charniak - 1997
256	Selective sampling using the query by committee algorithm, Machine Learning 28 - Freund, Seung, et al. - 1997
240	Inside-Outside Reestimation from Partially Bracketed Corpora - Pereira, Schabes - 1992
232	Trainable Grammars for Speech Recognition - Baker - 1979
93	Active Learning for Natural Language Parsing and Information Extraction - Thompson, Califf, et al. - 1999
70	Learning parse and translation decisions from examples with rich context - Hermjakob, Mooney - 1997
44	Minimizing Manual Annotation Cost in Supervised Training from Corpora - Engelson, Dagan - 1996
40	Lexicalized context-free grammars - Schabes, Waters - 1993
32	Parenthesis grammars - McNaughton - 1967
27	Selective Sampling for Example-Based Word Sense Disambiguation - Fujii, Tokunaga, et al. - 1998
24	Supervised grammar induction using training data with limited constituent information - Hwa - 1999
1	An empiric~al evaluation of probabilistic lexicaiized tree insertion grammars - Hwa - 1998