An Enhanced Framework of Subjective Logic for Semantic Document Analysis

Abstract

Unlike propositional logic which works on truth or falsity of statements, human judgements are subjective in nature having certain degree of uncertainty. Two different people will analyse and interpret a document in two different ways based on their background and current focus. In this paper we present an enhanced framework of subjective logic for automated single document analysis where each sentence in the document represents a proposition, and 'opinions' are constructed about this proposition to focus the degree of uncertainty associated with it. The 'opinion' about a sentence determines the significance of that sentence in a document. The input arguments are built automatically from a document in the form of evidence; then they are analyzed based on subjective logic parameters. Two different approaches are described here. The first utilises "bag of words" concept. However, this approach tends to miss the underlying semantic meanings of the context, so we further enhanced it into the latter approach which incorporates semantic information of the context, by extending the basic definitions of subjective logic.