Extracting Generalized Semantic Roles from Corpus

One of the oldest constructs of linguistic theory is semantic role. Automatic extraction of semantic roles in a sentence is a movement towards semantic processing of texts which has been the focus of attention in recent years. Extraction of semantic roles from a text contains some essential parts. Recognition of verb(s) of the sentence, recognition of noun phrases and their heads, and labeling the role of each phrase in the sentence as a semantic argument of verb are general parts of a system that does this task. There is a wide variety of definitions for semantic roles from verb specific roles to some general roles known as thematic roles, This paper focuses on a generalization of thematic roles called proto-roles or generalized semantic roles which includes two roles; actor and undergoer. In this paper we extract proto-roles in a Persian sentence exploiting POS tags. We use Peykareh as our input corpus and apply a rule based approach to extract actor and undergoer of verb(s).