Dong Song and Anoop Sarkar won the best paper award for their paper “Training Global Linear Models for Chinese Word Segmentation” in Canadian Conference on Artificial Intelligence-2009.
Paper Summary:
The paper gives a simple methodology for achieving state-of-the-art accuracy in Chinese word segmentation using global linear models. It explores different ideas for building Chinese word segmentator using global and local features and it employs reranking and beam search for enumerating all possible segmentations of the input. It further compares the average perceptron based approach with the Max-margin method for learning global feature weights and show that average perceptron based method achieve highest accuracy and faster segmentation for three different datasets.
