Opinion Stream Classification with Ensembles and Active leaRners - OSCAR
Final Report Abstract
Many people use social media to post opinions on almost any subject - events, products, topics. Institutions use these opinions to derive models. As opin- ions accumulate, though, changes occur and invalidate the models. Changes concern the general sentiment towards a subject and towards specific facets of this subject, as well as the words used to express sentiment. In OSCAR, we studied what people speak about, when they express opinions: we identified “words” that have a history and an impact, both of which are also expressed in words; we also identified “entities” behind the words; these are the objects about which people speak, e.g. products, ailments, treatments, as well as the subjects, namely the people themselves, e.g. as reviewers of products, as users of services, as patients in a self-help forum. We developed algorithms that monitor how the semantics of the words and the properties of the entities change over time. We tested our algorithms on streams of texts in social fora, on streams of opinions on products and on inter- actions of patients with eHealth platforms and mHealth apps. We found that our algorithms have great potential in fairness-aware machine learning and in the digital support of patient empowerment for patients with chronical diseases.
Publications
-
Large scale sentiment learning with limited labels. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1823–1832, 2017, ACM
Vasileios Iosifidis and Eirini Ntoutsi
-
Patient empowerment through summarization of discussion threads on treatments in a patient self-help forum. In: Nicos Maglaveras, Ioanna Chouvarda, and Paulo de Carvalho, editors, Precision Medicine Powered by pHealth and Connected Health, pages 229–233, Singapore, 2018. Springer Singapore
Sourabh Dandage, Johannes Huber, Atin Janki, Uli Niemann, Rüdiger u Pryss, Manfred Reichert, Steve Harrison, Markku Vessala, Winfried Schlee, Thomas Probst, and Myra Spiliopoulou
-
Sentiment classification over opinionated data streams through informed model adaptation. In International Conference on Theory and Practice of Digital Libraries,pages 369–381, 2017, Springer
Vasileios Iosifidis, Annina Oelschlager, and Eirini Ntoutsi
-
Active stream learning with an oracle of unknown availability for sentiment prediction. In Proceedings of the Workshop on Interactive Adaptive Learning (IAL’18) at ECML PKDD 2018, Dublin, Ireland, pages 36–47, 2018.
Elson Serrao and Myra Spiliopoulou
-
Learning under feature drifts in textual streams. In Proceedings of the 27th ACM International Conference on Information and Knowledge Management (CIKM’18), pages 527–536. 2018, ACM
Damianos P Melidis, Myra Spiliopoulou, and Eirini Ntoutsi
-
Predicting polarities of entity-centered documents with- out reading their contents. In Proceedings of the 33rd Annual ACM Symposium on Applied Computing (SAC’18), page 525–528, New York, NY, USA, 2018. ACM
Christian Beyer, Uli Niemann, Vishnu Unnikrishnan, Eirini Ntoutsi, and Myra Spiliopoulou
-
Entity-level stream classification: exploiting entity similarity to label the future observations referring to an entity. International Journal of Data Science and Analytics, Feb 2019
Vishnu Unnikrishnan, Christian Beyer, Pawel Matuszyk, Uli Niemann, Rüdiger Pryss, Winfried Schlee, Eirini Ntoutsi, and Myra Spiliopoulou
-
Exploiting entity information for stream classification over a stream of reviews. In Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing, SAC ’19, page 564–573, New York, NY, USA, 2019, ACM
Christian Beyer, Vishnu Unnikrishnan, Uli Niemann, Pawel Matuszyk, Eirini Ntoutsi, and Myra Spiliopoulou
-
Sentiment analysis on big sparse data streams with limited labels. Knowledge and Information Systems, pages 1–40, 2019
Vasileios Iosifidis and Eirini Ntoutsi
