ACUT: An Associative Classifier Approach to Unknown Word POS Tagging

The focus of this article is unknown word Part-of-Speech (POS) tagging. POS tagging which is one the fundamental requirements for intelligent text processing based on texts language. Therefore, this article firstly aims to provide a POS tagger with high accuracy for Persian language. The technique which is proposed by this article for handling unknown words is using a combination of a type of associative classifier along with a Hidden Markov Models (HMM) algorithm. Associative classification is a new classification approach integrating association mining and classification. The associative classifier used in this study is a type of associative classifiers that is innovated by this research. This kind of classifier not only uses sequence probability but also uses the CBA classifier. CBA first generates all the association rules with certain support and confidence thresholds as candidate rules. It then selects a small set of rules from them to form a classifier. When predicting the class label for an example, the best rule whose body is satisfied by the example is chosen for prediction. Based on the experimental results, the proposed algorithm can increase the accuracy of Persian unknown word POS tagging to 81.8 %. The total accuracy of proposed tagger is 98 % and its sentence accuracy is 63.1 %.