Ontological representation of texts, and its applicationsin text analysis

Abstract

For the management of a company, the need to know what people think of their products or services is becoming increasingly important in an increasingly competitive market. As the Internet can nearly be described as a digital mirror of events in the ”real“ world, being able to make sense of the semi structured nature of natural language texts published in this ubiquitous medium has received growing interest. The approach proposed in the thesis combines natural language processing and ontologies to represent elements in texts and relations between them. These relations are used to determine what terms describe what objects and what spot the descriptions put the objects in. Are the objects spoken of in a positive or negative manner? Throughout the report, it will be emphasized that ontologies can be applied to increase the ability of an automated agent to reason about texts. The more meta data that exist about a word or expression in a given context, the more deductions the agent will be able to make. One of the ontologies we developed allows the identification of statements that are considered either positive or negative relative to a given term.