This paper describes the Unitor system that participated to the SENTIment POLarity Classification task proposed in Evalita 2016. The system implements a classification workflow made of several Convolutional Neural Network classifiers, that generalize the linguistic information observed in the training tweets by considering also their context. Moreover, sentiment specific information is injected in the training process by using Polarity Lexicons automatically acquired through the automatic analysis of unlabeled collection of tweets. Unitor achieved the best results in the Subjectivity Classification sub-task, and it scored 2nd in the Polarity Classification sub-task, among about 25 different submissions.
Castellucci, G., Croce, D., Basili, R. (2016). Context-aware convolutional neural networks for twitter sentiment analysis in Italian. In CEUR Workshop Proceedings. CEUR-WS.
Context-aware convolutional neural networks for twitter sentiment analysis in Italian
CASTELLUCCI, GIUSEPPE;CROCE, DANILO;BASILI, ROBERTO
2016-01-01
Abstract
This paper describes the Unitor system that participated to the SENTIment POLarity Classification task proposed in Evalita 2016. The system implements a classification workflow made of several Convolutional Neural Network classifiers, that generalize the linguistic information observed in the training tweets by considering also their context. Moreover, sentiment specific information is injected in the training process by using Polarity Lexicons automatically acquired through the automatic analysis of unlabeled collection of tweets. Unitor achieved the best results in the Subjectivity Classification sub-task, and it scored 2nd in the Polarity Classification sub-task, among about 25 different submissions.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.