Modelling of Topic from Hindi Corpus using Word2Vec
Sabitra Sankalp Panigrahi, Narayan Panigrahi, Biswajit Paul · 2018
Detection of theme word or key word describing a collection of words is an important text processing method in natural language processing known as topic detection (TD). An accurate topic detection method depends upon goodness of topic modelling technique. There are several topic modelling techniques implemented successfully, some prominent names are LSA, LDA, Hierarchical Dirichlet Process, Non-Negative Matrix Factorization. Most topic modelling/detection techniques applied well for English corpus but very little work is available when it comes to Indian Languages. In this paper we have designed and applied a novel method for detection of topics from Hindi corpus. The proposed method discovers topics through clustering of semantic space; generated using word2vec word embedding. The results obtained are encouraging.