An Affinity Based Greedy Approach towards Chunking for Indian Languages

A robust chunker can drastically reduce the complexity of parsing of natural language text. Chunking for Indian languages require a novel approach because of the relatively unrestricted order of words within a word group. A computational framework for chunking based on valency theory and feature structures has been described here. The paper also draws an analogy of chunk formation in free word order languages with the bonding of atoms, radicals or molecules to form complex chemical structures. The unavailability of large annotated corpora forces one to adopt a statistical approach to achieve the task of word grouping for Indian language text. A chunker has been implemented for Bengali using this approach with considerably good accuracy.