Investigations into Distributional Semantics for Cognate Detection and Phylogenetics

This thesis investigates distributional semantics for cognate detection, false friends' detection and computational phylogenetics to present the insights drawn from our research, for 14 Indian languages pairs. Shared vocabulary facilitates second language learning and enables the computational models to perform cross-lingual learning for natural language processing (NLP) tasks. Distributional semantics aids NLP as it allows these models to understand natural languages. Our investigations use cross-lingual features to help detect cognates and false friends across languages. We also generate typological trees for Indian languages and, additionally, propose the division of the text into meaningful functional units which aid phylogenetic tree generation.