A hybrid approach for NER system for Scarce Resourced Language-URDU: Integrating n-gram with rules and gazetteers

We present a hybrid NER (Name Entity Recognition) system for Urdu script by integration of n-gram model (unigram and bigram), rules and gazetteers. We used prefix and suffix characters for rule construction instead of first name and last name lists or potential terms on the output list that is produced by n-gram model. Evaluation of the system is performed on two corpora, the IJCNLP NE (Named Entity) corpus and CRL NE corpus in Urdu text. The system achieved 92.65 and 87.6% using hybrid unigram and 92.47 and 86.83% using hybrid bigram on IJCNLP NE corpus and CRL NE corpus, respectively.