word2vec or JoBimText?
Nitin Ramrakhiyani, Sachin Sharad Pawar, Girish Keshav Palshikar · 2015
Exploration of distributional semantics for NLP tasks in Indian languages has been scarce. This work carries out a comparative analysis of two recent and high performing distributional semantics techniques namely word2vec and JoBimText. The task of lexical expansion of words in Hindi is considered for the analysis. A manual similarity assessment of the lexical expansions of words is employed for evaluation of the techniques. It can be observed that word2vec framework performs better than the JoBimText for various corpus sizes. Analysis of the results also presents insights on performance of the systems on various word types.