Graph Based Approach for Predicates Identification from Textual Data: A Case Study

Correct Identification of predicates like nouns, verbs and prepositions is a subtask of answer extraction in Machine Reading Comprehension system. With the help of NLP Tools and Stanford parser, the methodology is proposed for extraction of noun phrases, verbs and preposition. It is based on the concept of semantic relations associated with noun phrases, verbs and prepositions. It has been observed that the proposed methodology works with different genre of domains. This is implemented with different Reading Comprehension datasets. The proposed case study indicate that the methodology is generalized and can be applied to different datasets. The knowledgebase generated with predicates identification can be used in various application which are based on information extraction. It ensures the concept of generalization and transfer domain. This work highlights the problems identified in predicates identification when applied to different datasets. It also suggests the solutions to overcome some of the problems, which indicates the future research.