SemEval-2007 Task 12: Turkish Lexical Sample Task

Proceedings of the 4th International Workshop on Semantic Evaluations (SemEval-2007), pages 59–63, Prague, June 2007. c⃝2007 Association for Computational Linguistics SemEval-2007 Task 12: Turkish Lexical Sample Task Zeynep Orhan Department of Computer Engineering, Fatih University 34500, Büyükçekmece, Istanbul, Turkey zorhan@fatih.edu.tr Emine Çelik Department of Computer Engineering, Fatih University 34500, Büyükçekmece, Istanbul, Turkey eminemm@gmail.com Neslihan Demirgüç Department of Computer Engineering, Fatih University 34500, Büyükçekmece, Istanbul, Turkey nesli_han@hotmail.com Abstract This paper presents the task definition, re- sources, and the single participant system for Task 12: Turkish Lexical Sample Task (TLST), which was organized in the Se- mEval-2007 evaluation exercise. The methodology followed for developing the specific linguistic resources necessary for the task has been described in this context. A language-specific feature set was defined for Turkish. TLST consists of three pieces of data: The dictionary, the training data, and the evaluation data. Finally, a single system that utilizes a simple statistical method was submitted for the task and evaluated. 1 Introduction Effective parameters for word sense disambigua- tion (WSD) may vary for different languages and word types. Although, some parameters are com- mon in many languages, some others may be lan- guage specific. Turkish is an interesting language that deserves being examined semantically. Turk- ish is based upon suffixation, which differentiates it sharply from the majority of European languages, and many others. Like all Turkic languages, Turk- ish is agglutinative, that is, grammatical functions are indicated by adding various suffixes to stems. Turkish has a SOV (Subject-Object-Verb) sentence structure but other orders are possible under certain discourse situations. As a SOV language where objects precede the verb, Turkish has postpositions rather than prepositions, and relative clauses that precede the verb. Turkish, as a widely-spoken lan- guage, is appropriate for semantic researches. TLST utilizes some resources that are explained in Section 2-5. In Section 6 evaluation of the sys- tem is provided. In section 7 some concluding re- marks and future work are discussed. 2 Corpus Lesser studied languages, such as Turkish suffer from the lack of wide coverage electronic re- sources or other language processing tools like on- tologies, dictionaries, morphological analyzers, parsers etc. There are some projects for providing data for NLP applications in Turkish like METU Corpus Project (Oflazer et al., 2003). It has two parts, the main corpus and the treebank that con- sists of parsed, morphologically analyzed and dis- ambiguated sentences selected from the main cor- pus, respectively. The sentences are given in XML format and provide many syntactic features that can be helpful for WSD. This corpus and treebank can be used for academic purposes by contract. The texts in main corpus have been taken from different types of Turkish written texts published in 1990 and afterwards. It has about two million words. It includes 999 written texts taken from 201 books, 87 papers and news from 3 different Turk- ish daily newspapers. XML and Text Encoding Initiative (TEI) style annotation have been used. The distribution of the texts in the Treebank is similar to the main …

Content truncated to 500 words