Automatic Identification of Bengali Noun-Noun Compounds Using Random Forest

This paper presents a supervised machine learning approach that uses a machine learning algorithm called Random Forest for recognition of Bengali noun-noun compounds as multiword expression (MWE) from Bengali corpus. Our proposed approach to MWE recognition has two steps: (1) extraction of candidate multi-word expressions using Chunk information and various heuristic rules and (2) training the machine learning algorithm to recognize a candidate multi-word expression as Multi-word expression or not. A variety of association measures, syntactic and linguistic clues are used as features for identifying MWEs. The proposed system is tested on a Bengali corpus for identifying noun-noun compound MWEs from the corpus.