A language independent framework for representing event semantics in modern persian

Semantics, the study of meaning, is an important and active area of research. It allows computational and logical understanding of the meaning entailed by language. In this dissertation we present SERF, Semantic Event Representation Framework, a novel system for representing event semantics. Our framework is inspired by frame semantics, casting prototypical events as frames with semantic roles, or "slots", to be filled. Unlike most other representational means in this domain, our framework does not depend on natural language definitions. Instead, it uses semantic relations to symbolize the associations among the roles of an event, combining into semantic networks for each event prototype. These semantic networks contain relations that hold within the context of a particular event prototype. Another original feature is the ability to represent semantics that are dependent upon temporal contexts—that is, identifying relations that hold initially at the outset of an event, those which hold terminally, and how that change occurs. We propose the use of a small, fixed number of semantic relations and entities for representing event prototypes in order to facilitate greater language independence. The advantage of language independent features is two-fold. First, it facilitates the rapid prototyping of semantic resources for new languages, especially important for languages underrepresented by natural language processing resources, so-called "minority languages". Second, it provides a mechanism for taking semantic features that are easy to detect in certain languages and mapping them onto other languages where their evidence is more obscure. We thus introduce Persian SERF Net, an implementation of our framework for Modern Persian that is incorporated into a WordNet-like resource. Within our framework, the semantic network that represents a particular event prototype is evoked by a verb sense, and as such is associated with a corresponding verb synset in our resource. That is, each verb synset contains a semantic network that represents its associated event semantics. Due to particular linguistic features of Persian, we are able to include additional relations in our implementation that span parts-of-speech. Finally, we demonstrate some applications of the Persian resource to two semantic tasks: interpretation of metaphor and inference of meaning through implicature.