Approach for multiword expression recognition & annotation in urdu corpora
Vaishali Gupta, Nisheeth Joshi, Iti Mathur · 2017
Multiword Expression (MWE) is an open challenging task in NLP because of its compositional behavior. In this paper, we have presented the classification of multiword expressions and also designed some MWE tags for annotation of Urdu corpora. If MWE annotation is proper then it can be accurately processed through parsing phase and can improve the accuracy of machine translator. For annotation of MWEs, initially extracted the compound type MWE on the basis of automatic approach and also extracted different type of MWEs using semiautomatic and manual approach. These extracted MWEs are pre-processed and after that got the correct multiword expressions. Using these MWEs, we have annotated the Urdu corpora of 50,000 sentences.