Tukish Part of Speech Tagging with Deep Learning Methods
Project no. 177 · Status: Kapanmış · 2018
Summary
Natural language processing is one of the important areas of artificial intelligence. In addition to basic studies such as morphological analysis, part of speech tagging, studies in the field of natural language processing include language production, speech analysis, automatic translation between languages, information retrieval, text correction, automatic indexing, speech understanding. It can be said that natural language processing studies are structured on each other or increase each other's performance. It has been shown in many studies that especially part of speech tagging increases the performance rates in areas such as information retrieval and language understanding. In this project, it is aimed to solve the word type detection problem, which has previously been provided with a high performance rate with rule-based and statistical methods on Turkish, with a deep learning approach. To our knowledge, there is no other study in Turkish that deals with part of speech tagging problem with deep learning methods. In the project, deep learning methods have been used together with the word embeddings. SENNA (Semantic / Syntactic Extraction using a Neural Network Architecture) tool was used in the deep learning processes in the research. As a result of the performance tests in of our setting, a success rate of 83.09% has been achieved. Although this performance rate is relatively low compared to the performance rates obtained by different methods, it can be considered as a good starting point for future studies.