Words are not Equal: Graded Weighting Model for Document Vectors

In recent years distributional semantics or vector models for words have been proposed to capture both the syntactic and semantic similarities between words. Such vetors may be obtained for words as used in a large corpus or in a given domain. Since these are language free models and can be obtained in an unsupervised manner they are of interest for under-resourced languages such as Hindi. We start with an overview which shows that a reasonable measure of semantic similarity in Hindi seems to be captured by a word vector map.