Turkish named entity recognition with deep learning

Abstract

Named Entity Recognition (NER) is an important task in Natural Language Processing, Data Mining and Information Extraction areas since 1990's. While NER is a succesfully solved problem in English, it is still a hot topic in agglutinative languages like Turkish, Czech, Finnish languages. With the scope of this study we focus on Bidirectional Long Short-Term Memory (BLSTM) neural network models to solve NER problem. We suggest a succesful implementation of Deep Bidirectional Long Short Term Memory (DBLSTM) which reaches %93.69 F1 score, which is state-of-the-art result for Named Entity Recognition in Turkish.