Forming Semantics of Turkish Texts in an Application Area
Semantics is the hardest and most important part of the NLPs three level structure: Morphology, Syntax, Semantics. It is the top level and other levels outputs are used. In the semantics level, the sentences of natural languages are translated into semantic formulas; which can further be used for applications like machine translation and human-machine conversation.
In this study, several methods are developed to derive semantic representations of Turkish sentences. Firstly, to start with a small subset of Turkish is chosen: the recipes. The sentences in the recipes have some characteristics that make syntactic rules and semantic formulas simpler. Despite this simplicity those sentences contain several complex noun and adverbial phrases. Especially different adverb types cause problems in semantic representation.
The main application developed for this study reads the recipe files, and outputs the semantic representations (or inter-lingua formulas) of the sentences. Another program is implemented to translate some of these sentences into English, to check the validity of the semantic outputs of the main application.