Reasoning with Text Annotations

With the emerging need for automation of business processes and the advent of semantic web it has become necessary that digital contents should be expressed not only in natural language, but also in a form that can be understood, interpreted and used by software agents, thus permitting them to find, share and integrate information more easily. Thus, Knowledge Representation and automated reasoning has been an interesting field of research for recent years, which is greatly supported by the availability of various logical languages underpinned with well defined syntax and semantics. Description logic (DL), which is also the foundation of standardized web ontology language OWL, has become a highly used de-facto standard for knowledge representation. DL reasoners can infer and detect logical contradictions in the ontologies specified in a certain web ontology language, such as OWL. Highly optimized DL reasoners like Pellet, Hermit, Fact++ can be used for reasoning over knowledge base represented using classical DL. Annotated corpus with various annotations viz semantic, syntactic and pragmatic are important features for text-based applications, but are equally challenging and complex to be achieved, maintained and be used in an integrated way. So, a framework in which various linguistic annotations, produced by different ad hoc approaches and domain knowledge maintained by experts are integrated is useful for enrichment of the corpus. With standard representation formalism the annotated documents can be made suitable for automated reasoning. This formalism not only new annotations can be inferred from the existing ones but the same formalism and framework can be used for checking the correctness and completeness of annotations. In this thesis the underlying knowledge representation formalism for representation of annotations is analyzed and solved by studying various logical languages including higher order logics, which accounts for cooperation of the different sorts of knowledge. The annotations are interpreted in various ways to handle the different needs to achieve varying useful results. At the same time different paradigms for reasoning with annotations are discussed along with necessary algorithms for dealing with such cases. An integrated framework is developed, the prototype version of which demonstrates the use of reasoning taking into account of both linguistic and semantic annotations, to detect conflicting annotations, new annotations and overall consistency of annotations. Thus, the research work that we have undertaken helped us to achieve an integrated knowledge representation and reasoning framework for dealing with annotated text corpus,supported by existing cutting age semantic web technologies. This thesis, also paves a path towards a metric to evaluate annotation i.e. detect logically inconsistent information, without depending on a golden standard, but by making use of the axioms expressed in ontologies.