Investigation of Aspect Based Turkish Sentiment Analysis Subtasks – Identification of Aspect Term, Aspect Category And Sentiment Polarity

Sentiment analysis studies conducted traditionally at document or sentence level have been moved to a new level with the emergence of aspect based sentiment analysis studies. Aspect-based sentiment analysis can be briefly defined as the detection of different opinions contained within a text together with the target entities to which they relate. Current definitions describe aspect based sentiment analysis as a gradual task aiming to identify opinion tuples represented by three main fields (target term, target category, sentiment class). This article presents our investigations on aspect based Turkish sentiment analysis. The work carried out in this article is designed by following ABSA 2016 competition tasks (1- Aspect category identification, 2- Aspect term identification, 3- Identification of aspect category and aspect term together and 4- sentiment category classification) and evaluated on the Turkish restaurant reviews dataset provided in the same event. For the first three tasks, a sequence labeling algorithm (based on conditional random fields (CRF)) which uses word vectors and natural language processing outputs (word and sentence analyses) is proposed and shown to solve these three tasks in one step. Experimental results show that the proposed system achieves the highest performances for these tasks: 66.7% F1-score for aspect category identification, 53.2% F1-score for aspect term identification, 46.7% F1-score for both aspect category and aspect term at the same time. Additionally, a linear classification method based on feature selection from positionally and syntactically neighboring tokens is proposed for sentiment category classification task and shown to perform as the best constrained system reported in the literature with 76.1% F1-score