Name finding from free text using HMMS

Abstract

In this paper we design and implement a HMM to solve the Named Entity (NE) task. The NE task is used to recognize and classify names, dates, and numerical quantities, although we restrict our model to names and dates. Our results were compared with other Named Entity Recognition (NER) systems that evaluated their model based on data from the Sixth and Seventh Message Understanding Conferences (MUC-6 and MUC-7), and other domain data. We show the importance of the amount of training data available and the importance of back-off models and smoothing. Although our algorithm was not tested on a large amount of training data, performance was quite good. We evaluate our results and compare them to other related work, (Valarkos A. et al, 2003) (Bikel D. et al, 1999) (Masterson D., 1999).