NEUROSENT-PDI at SemEval-2018 Task 1: Leveraging a Multi-Domain Sentiment Model for Inferring Polarity in Micro-blog Text

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 102–108 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics NEUROSENT-PDI at SemEval-2018 Task 1: Leveraging a Multi-Domain Sentiment Model for Inferring Polarity in Micro-blog Text Mauro Dragoni Fondazione Bruno Kessler Via Sommarive 18 Povo, Trento, Italy dragoni@fbk.eu Abstract This paper describes the NeuroSent system that participated in SemEval 2018 Task 1. Our system takes a supervised approach that builds on neural networks and word embeddings. Word embeddings were built by starting from a repository of user generated reviews. Thus, they are specific for sentiment analysis tasks. Then, tweets are converted in the correspond- ing vector representation and given as input to the neural network with the aim of learning the different semantics contained in each emo- tion taken into account by the SemEval task. The output layer has been adapted based on the characteristics of each subtask. Prelimi- nary results obtained on the provided training set are encouraging for pursuing the investiga- tion into this direction. 1 Introduction Sentiment Analysis is a natural language process- ing (NLP) task (Dragoni et al., 2015) which aims at classifying documents according to the opin- ion expressed about a given subject (Federici and Dragoni, 2016a,b). Many works available in the literature address the sentiment analysis problem without distinguishing domain specific informa- tion of documents when sentiment models are built. The necessity of investigating this prob- lem from a multi-domain perspective is led by the different influence that a term might have in dif- ferent contexts. The idea of adapting terms po- larity to different domains emerged only in the last decade (Blitzer et al., 2007; Dragoni and Petrucci, 2017). Multi-domain sentiment analy- sis approaches discussed in the literature focus on building models for transferring information be- tween pairs of domains (Dragoni, 2015; Petrucci and Dragoni, 2015). While on the one hand such approaches allow to propagate specific domain in- formation to others, their drawback is the neces- sity of building new transfer models every time a new domain has to be analyzed. Thus, such ap- proaches do not have a great generalization capa- bility of analyzing texts, because transfer models are limited to the N domains used for building the models. The NeuroSent tool applied in SemEval 2018 Task 1 (Mohammad et al., 2018) leverages on the following pillars: (i) the use of word embeddings for representing each word contained in raw sen- tences; (ii) the word embeddings are generated from an opinion-based corpus instead of a general purpose one (like news or Wikipedia); (iii) the de- sign of a deep learning technique exploiting the generated word embeddings for training the sen- timent model; and (iv) the use of multiple output layers for combining domain overlap scores with domain-specific polarity predictions. The last point enables the exploitation of lin- guistic overlaps between domains, which can be considered one of the pivotal assets of our ap- proach. This way, the overall polarity of a doc- ument is computed by aggregating, for …

Content truncated to 500 words