Part-of-Speech Tagging with Two Sequential Transducers

The article presents a method of constructing and applying a cascade consisting of a leftand a right-sequential finite-state transducer, and , for part-of-speech disambiguation. In the process of POS tagging, every word is first assigned a unique ambiguity class that represents the set of alternative tags that this word can occur with. The sequence of the ambiguity classes of all words of one sentence is then mapped by to a sequence of reduced ambiguity classes where some of the less likely tags are removed. That sequence is finally mapped by to a sequence of single tags. Compared to a Hidden Markov model tagger, this transducer cascade has the advantage of significantly higher processing speed, but at the cost of slightly lower accuracy. Applications such as Information Retrieval, where the speed can be more important than accuracy, could benefit from this approach.