The Szeged Treebank Project

The major aim of the Szeged Treebank project was to create a high-quality database of syntactic structures for Hungarian that can serve as a golden standard to further research in linguistics and computational language processing. The treebank currently contains full syntactic parsing of about 82,000 sentences (1.2 million words), which is the result of accurate manual annotation. Inspired by the research results of the Penn Treebank [6] and several other treebank projects [1,2,3,5,7], our research group set out to create a golden standard treebank for Hungarian, containing reliable syntactic annotation of texts. Project work contained the selection and adjustment of the theory used for syntactic analysis, the design of the annotation methodology, the adaptation of the available tag-sets to Hungarian, automated preprocessing, manual validation and correction, and experiments with machine learning methods for automated parsing. The proposed poster is to presents an overview of the Szeged Treebank initiative and its results to date. Ideally, the treebank should contain samples of all the syntactic structures of the language, therefore, it serves as a reference for future corpus and treebank developments, grammar extraction and other linguistic research. It also serves as a reliable test suite for different NLP applications, as well as a basis for the development of computational methods for both shallow and deep syntactic parsing, and information extraction. Well-defined methods or elaborate theoretical foundations for the automated syntactic analysis of Hungarian texts were lacking at the start of the project. For this reason, novelty of the project work lies in the design of a practical approach for syntactic annotation of Hungarian natural language sentences.