A framework for classifying semantic relationships

Recently, the NLP community has shown a renewed interest in lexical semantics in the extent of automatic recognition of semantic relationships between pairs of words in text. Lexical semantics has become increasingly important in many natural language applications, this approach to semantics is concerned with psychological facts associated with meaning of words and how these words can be connected in semantic relations to build ontologies that provide a shared vocabulary to model a specified domain. And represent a structural framework for organizing information across fields of Artificial Intelligence (AI), Semantic Web, systems engineering and information architecture. But current systems mainly concentrate on classification of semantic relations rather than to give solutions for how these relations can be created [14]. At the same time, systems that do provide methods for creating the relations tend to ignore the context in which the conceptual relationships occur. Furthermore, methods that address semantic (non-taxonomic) relations are yet to come up with widely accepted ways of enhancing the process of classifying and extracting semantic relations. In this research we will focus on the learning of semantic relations patterns between word meanings by taking into consideration the surrounding context in the general domain. We will first generate semantic patterns in domain independent environment depending on previous specific semantic information, and a set of input examples. Our case of study will be causation relations. Then these patterns will classify causation in general domain texts taking into consideration the context of the relations, and then the classified relations will be used to learn new causation semantic patterns.