Beyza Eken and A. Cuneyd Tantug, Istanbul Technical University, Turkey

Named entity recognition (NER) is one of the well-studied sub-branch of natural language processing (NLP). State of the art NER systems give highly accurate results in domain of formal texts. With the expansion of microblog sites and social media, this informal text domain has become a new trend in NLP studies. Recent works has shown, social media texts are hard to process and the performance of the current systems substantially decrease when switched to this domain. We give our experience in improving named entity recognition on informal social media texts for the case of tweets.

Named Entity Recognition, Conditional Random Fields, Informal Domain, Tweet, Turkish