Can We Mimic Human Pragmatics Knowledge into Computational Lex- icon?

So far Natural Language Processing (NLP) research patronized much of manually augmented lexicon resources such as WordNet. But the small set of semantic relations like Hypernym, Holonym, Meronym and Synonym etc are very narrow to capture the wide variations human pragmatics knowledge i.e. a news article containing the themes, “Iraq”, “Al-Qaeda”, “9/11” and “Osama Bin Laden” might suggest the topic related to “terrorism”. But no such information could be retrieved from available lexicon resources. SemanticNet is the attempt to capture wide context dependent semantic inference among various themes which human being perceives in their pragmatic knowledge, learnt by day to day cognitive interactions with physical world surrounded by them. One such effort towards capturing the human common sense is ConceptNet (Liu and Singh, 2004) for English. We extend our vision over the human common sense to human pragmatics and proposed semantic relations for every pair of lexicons cannot be defined by fixed number of certain semantic relation labels and thus we formulated a probabilistic score based technique. SemanticNet is a semantic network of lexicons to hold human pragmatic knowledge. Contextual semantic affinity inference in SemanticNet could be calculated by network distance. SemanticNet presently developed for Bengali language.