Automatic Correction of Disfluencies in Tamil Disfluent Text: A Rule-Based Approach
摘要
This research paper presents a comprehensive approach on the automatic correction of four types of disfluencies in Tamil disfluent text. Correction of Interjection, Prolongation, whole-word repetition and part-word repetition disfluencies is the focus of this research work. The disfluent Tamil text dataset used in this study comprises 5000 synthesized sentences. The proposed Tamil Text Correction Tool (TTCT) applies the non-deterministic finite automaton simulation to correct disfluencies by backtracking. The proposed framework explores various possibilities and backtracks when a path fails. Thereby, the aforementioned disfluencies were automatically corrected and presented as fluent Tamil sentences. TTCT achieves cent percent overall detection accuracy and in correcting individual disfluency types.