Paper: Improving Sequence Segmentation Learning By Predicting Trigrams

ACL ID W05-0611
Title Improving Sequence Segmentation Learning By Predicting Trigrams
Venue International Conference on Computational Natural Language Learning
Session Main Conference
Year 2005

Symbolic machine-learning classifiers are known to suffer from near-sightedness when performing sequence segmentation (chunking) tasks in natural language pro- cessing: without special architectural ad- ditions they are oblivious of the decisions they made earlier when making new ones. We introduce a new pointwise-prediction single-classifier method that predicts tri- grams of class labels on the basis of win- dowed input sequences, and uses a simple voting mechanism to decide on the labels in the final output sequence. We apply the method to maximum-entropy, sparse- winnow, and memory-based classifiers us- ing three different sentence-level chunk- ing tasks, and show that the method is able to boost generalization performance in most experiments, attaining error reduc- tions of up to 51%....