🧑‍🔬 Yee Seng Chan

Yee Seng Chan works on word sense disambiguation and its role in statistical machine translation. The papers here ask two linked questions. The first is how to obtain sense-labelled training examples without costly manual annotation. The second is whether resolving lexical ambiguity actually yields better translations. A recurring method is to treat the translations found in word-aligned English–Chinese parallel corpora as the labels for senses. The distinctions a system learns are then the ones that matter for choosing words in another language.

The most cited paper, Word Sense Disambiguation Improves Statistical Machine Translation (2007), responded to earlier studies that disagreed on this point. It plugged a strong disambiguation classifier into Hiero, a hierarchical phrase-based system, and tested it on a full Chinese-to-English task. The classifier's output enters only as extra feature scores on rules the system already has, not as new rules. A second, count-like feature offsets the penalty that a sparsely firing probability feature would otherwise impose. The earlier study, Exploiting Parallel Texts for Word Sense Disambiguation (2003), was written with Hwee Tou Ng and Bin Wang. It examined whether aligned bilingual text could stand in for hand-tagged examples. English noun senses from WordNet that share a Chinese translation were merged, and the aligned English contexts were used to train a naive Bayes classifier.

The 2007 paper is also the newest of Chan's papers here. It shifts the emphasis from gathering training data for disambiguation to putting disambiguation to work inside a translation system. Chan is currently affiliated with RTX (United States). Earlier bylines point to the University of Illinois Urbana-Champaign and, through the 2000s, the National University of Singapore. Co-authors on these papers include Hwee Tou Ng, Bin Wang and David Chiang. The topics are word sense disambiguation and machine translation.

Yee Seng Chan has 2 papers in our library, in 2003–2007, most of them on Word sense disambiguation, Machine translation; often with Hwee Tou Ng, Bin Wang, David Chiang.