Kb context switching algorithm for nlp

Abstract

Those who develop Natural Language Processing (NLP) systems sometimes find it convenient to develop several representations of knowledge. Especially in question and answer generating machines it is more logical to have several knowledge bases (KB) to answer each specific types of questions. This paper discusses a statistical learning based algorithm to port a specific question to a desired KB. The algorithm allows selection of KBs based on previously learnt patterns. Due to probabilistic parsing and POS (Part of Speech) tagging makes this algorithm much suitable for short questions or short input sentences.