A Supervised Part-Of-Speech Tagger for the Greek Language of the Social Web

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3861–3867 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3861 A Supervised Part-Of-Speech Tagger for the Greek Language of the Social Web Maria Nefeli Nikiforos, Katia Lida Kermanidis Department of Informatics, Ionian University Corfu, Greece c19niki@ionio.gr, kerman@ionio.gr Abstract The increasing volume of communication via microblogging messages on social networks has created the need for efficient Natural Language Processing (NLP) tools, especially for unstructured text processing. Extracting information from unstructured social text is one of the most demanding NLP tasks. This paper presents the first part-of-speech tagged data set of social text in Greek, as well as the first supervised part-of-speech tagger developed for such data sets. Keywords: supervised part-of-speech tagging, social web language, Greek 1. Introduction The increasing volume of communication via microblog- ging messages on social networks (such as Facebook, Twit- ter, Blogs and YouTube) has created the need for devel- oping efficient Natural Language Processing (NLP) tools, especially for unstructured text processing. It has been ob- served that the performance of NLP tools developed for structured text processing is significantly reduced when used in unstructured microblogging text (Nand et al., 2014). Therefore, there is a growing demand for information ex- traction tools from unstructured social text for business in- telligence, security, programming etc. However, extracting information in this context is one of the most demanding NLP tasks due to the unusual structure of the text (Nand et al., 2014). As a result, it is necessary to either adapt existing methods for implementation in this context, or to develop new methods that perform well on unstructured microblog- ging text. Part-of-speech tagging (POS tagging) is a fundamental part of NLP (Gimpel et al., 2010). A robust POS tagging tool has an important role in most NLP problems and appli- cations, such as syntactic and semantic analysis, machine translation and sentiment analysis (Bach et al., 2018; Liu et al., 2012). The main challenges in POS tagging are: a. cre- ating the tag set, based on which each word will be labeled with a POS tag, which needs to include the specifics of the social network context, b. labeling ambiguous words with the correct tag, and c. creating a labeled data set of suffi- cient size for the best possible results in machine learning. Despite the development of certain NLP tools for con- ventional text in Greek (Papageorgiou et al., 2000; Peta- sis et al., 2001; Sgarbas et al., 2001), there are not any POS taggers for social text in Greek. The contribu- tion of our research is: a. creating the first annotated data set for Greek social text (2,405 tweets or 31,697 to- kens), b. creating the first tag set (22 different tags) in- c.cluding special tags for Twitter language specifics, and developing the first supervised POS tagger for such data with significantly high performance (accuracy up to 99.87%). The data set is available for research pur- poses at this address: https://hilab.di.ionio. gr/index.php/en/datasets/ …

Content truncated to 500 words