Token identification using HMM and PPM models

Abstract

Hidden markov models (HMMs) and prediction by partial matching models (PPM) have been successfully used in language processing tasks including learning-based token identification. Most of the existing systems are domain- and language-dependent. The power of retargetability and applicability of these systems is limited. This paper investigates the effect of the combination of HMMs and PPM on token identification. We implement a system that bridges the two well known methods through words new to the identification model. The system is fully domain- and language-independent. No changes of code are necessary when applying to other domains or languages. The only required input of the system is an annotated corpus. The system has been tested on two corpora and achieved an overall F-measure of 69.02% for TCC, and 76.59% for BIB. Although the performance is not as good as that obtained from a system with language-dependent components, our proposed system has power to deal with large scope of domain- and language-independent problem. Identification of date has the best result, 73% and 92% of correct tokens are identified for two corpora respectively. The system also performs reasonably well on people s name with correct tokens of 68% for TCC, and 76% for BIB.