Bidirectional Attentional LSTM for Aspect Based Sentiment Analysis on Italian
Giancarlo Nicola · Accademia University Press eBooks · 2018
This paper describes the SentITA system that participated to the ABSITA task proposed in Evalita 2018. The system is based on a Bidirectional Long Short Term Memory network with attention that exploits word embeddings and sentiment specific polarity embeddings. The model also leverages grammatical information from POS tagging and NER tagging. The system participated in both the Aspect Category Detection (ACD) and Aspect Category Polarity (ACP) tasks achieving the 5th place in the ACD task and the 2nd in the ACD task.