Sentiment analysis (SA) concerns identifying opinions or attitudes in text, and is often approached as the task of detecting the overall negative or positive polarity of a document. So-called aspect-based sentiment analysis (ABSA) aims at a more fine-grained analysis; identifying individual polar expressions, their corresponding targets or aspects, and sometimes even the holder or source of the expressed opinion. The SANT project at LTG recently released an annotated dataset for fine-grained SA for Norwegian; NoReCfine. The aim of this MSc project will be to train and evaluate different neural architectures for fine-grained sentiment analysis in Norwegian.
