Enza Messina is a Professor in Operations Research at the Department of Informatics Systems and Communications, University of Milano-Bicocca, Italy, where she leads the research Laboratory MIND (Models in Decision Making and Data Analysis). She holds a PhD in Computational Mathematics and Operations Research from the University of Milano. Her research activity is mainly focused on decision models under uncertainty and more recently on statistical relational models for data analysis and knowledge extraction. In particular, she developed relational classification and clustering models that find applications in different domains such as systems biology, e-justice, text mining and social network analysis. She is a co-founder of Sharper Analytics, a spin-off from the University of Milano Bicocca.
Deep Learning for Sentiment Analysis
We show how deep learning and ensemble methods can successfully address challenging problems arising in sentiment analysis such as irony detection or domain adaptation. In particular, we propose an unsupervised framework for domain-independent irony detection built upon an existing probabilistic topic model initially introduced for sentiment analysis purposes. Moreover, in order to improve its generalization abilities, we apply Word Embeddings to obtain domain-aware ironic orientation of words. The acquisition of cross-domain high level feature representations through word embeddings combined with the generalization capability of ensemble methods can also be used for addressing the problem of domain adaptation also in the scenario where the testing target domain is completely unlabelled.