A hybrid approach for urdu sentence boundary disambiguation

Sentence boundary identification is a preliminary step for preparing a text document for Natural Language Processing tasks, e.g., machine translation, POS tagging, text summarization and etc. We present a hybrid approach for Urdu sentence boundary disambiguation comprising of unigram statistical model and rule based algorithm. After implementing this approach, we obtained 99.48% precision, 86.35% recall and 92.45% F1-Measure while keeping training and testing data different from each other, and with same training and testing data, we obtained 99.36% precision, 96.45% recall and 97.89% F1-Measure.