Evaluation of Named Entity Features for Punjabi Language

Named entity recognition is a task to identify and classify the words in the given text to some predetermined categories like Organization, Location, Time, Number, Person etc. In this paper, NER system for Punjabi language has been evaluated on various combinations of features including context word window feature of 3, 5 and 7, various digit features, infrequent word and length of word features. After evaluation it has been found that the feature set comprising of word window 5, digit features, Infrequent word and Length of word feature has confirmed the highest f-score value of 87.46% with Precision and Recall values of 90.99% and 84.19% respectively. It has been realized that the feature set consisting of all language independent features give better results with word window 5 as compared to word window 3 and 7.