Malayalam Clause Boundary Identifier: Annotation and Evaluation
Sobha Lalitha Devi, S. Lakshmi · 2013
Clause boundary identification has a significant role in improving the performance of different practical NLP systems. In this paper we have dealt with automatically identifying various types of clausal structures in Malayalam, a Dravidian language. The clausal sentences were collected from tourism and health domain available in the Web. We discuss about the annotation schema and the inter-annotators agreement for various clauses and also the automatic identification of clause boundaries using CRFs a Machine learning approach. To smooth the errors obtained from the CRFs tagging, we have used linguistic rules. For Inter-annotators agreement we have used kappa coefficient as the agreement statistic. The evaluation gave encouraging result. 1