Time and place: November 29th, 9:30, room T5 (CS Otaniemi) and zoom

Speaker: Grigorios Tsoumakas,  Aristotle University of Thessaloniki (AUTH)

Title: Neural Abstractive Summarization: Methods and Applications

Abstract: This talk reviews past and recent work of our team on the topic of neural abstractive summarization. We will first present our divide-and-conquer approach for dealing with long documents and its application to summarizing scientific articles [1]. We will then discuss Bayesian active summarization, our approach to combining active learning with state-of-the-art summarization models [2, 3]. Next, we will share our methods towards controlling the output of summarization models given a particular context, such as a topic, along with our corresponding evaluation metric [4]. Finally, we will present applications in healthcare and finance [5, 6, 7].

References

[1] Gidiotis, A., & Tsoumakas, G. (2020). A Divide-and-Conquer Approach to the Summarization of Long Documents. IEEE/ACM Transactions on Audio Speech and Language Processing, 28, 3029–3040. https://doi.org/10.1109/TASLP.2020.3037401

[2] Gidiotis, A., & Tsoumakas, G. (2022). Should We Trust This Summary? Bayesian Abstractive Summarization to The Rescue. Proceedings of the Annual Meeting of the Association for Computational Linguistics, 4119–4131. https://doi.org/10.18653/V1/2022.FINDINGS-ACL.325

[3] Gidiotis, A., & Tsoumakas, G. (2024). Bayesian active summarization. Computer Speech & Language, 83, 101553. https://doi.org/10.1016/J.CSL.2023.101553

[4] Passali, T., & Tsoumakas, G. (2024) Topic-Controllable Summarization: Topic-Aware Evaluation and Transformer Methods. Proceedings of LREC COLING 2024. https://aclanthology.org/2024.lrec-main.1415/

[5] Giannouris, P., Myridis, T., Passali, T., & Tsoumakas G. (2024) Plain Language Summarization of Clinical Trials. Proceedings of LREC COLING 2024 DeTermIt! Workshop on Evaluating Text Difficulty in a Multilingual Context, pages 60–67, Torino, Italia. https://aclanthology.org/2024.determit-1.6/

[6] Stefanou, L, Passali, T. & Tsoumakas, G. (2024) AUTH at BioLaySumm 2024: Bringing Scientific Content to Kids. Proceedings of the 23rd Workshop on Biomedical Natural Language Processing, pages 793–803, Bangkok, Thailand. Association for Computational Linguistics. https://aclanthology.org/2024.bionlp-1.73/

[7] Passali, T., Gidiotis, A., Chatzikyriakidis, E., & Tsoumakas, G. (2021). Towards Human-Centered Summarization: A Case Study on Financial News. Bridging Human-Computer Interaction and Natural Language Processing, HCINLP 2021 - Proceedings of the 1st Workshop. https://aclanthology.org/2021.hcinlp-1.4/

Bio: Dr. Grigorios Tsoumakas received a degree in Computer Science from the Aristotle University of Thessaloniki (AUTH), Greece, in 1999, an MSc in Artificial Intelligence from the University of Edinburgh, United Kingdom, in 2000 and a PhD in Computer Science from AUTH in 2005. He is a Professor of Machine Learning and Knowledge Discovery at the School of Informatics of AUTH since 2024, where he has also served as Associate Professor (2020-2024), Assistant Professor (2013 – 2020) and Lecturer (2007 – 2013). Since 2024, he also serves as an Affiliate Researcher at Archimedes/RC Athena, Greece. In addition, he is co-founder and chief scientific officer at Medoid AI, a spin-off company of AUTH established in 2019, developing custom AI solutions based on cutting-edge Machine Learning technology. Dr. Tsoumakas is a senior member of ACM and IEEE. His research expertise focuses on supervised learning (ensemble methods, multi-target prediction, interpretablity) and natural language processing (semantic indexing, keyphrase extraction, summarization). He has published more than 150 research papers and according to Google Scholar he has more than 19.000 citations and an h-index of 52. His honors include receiving the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) 10-Year Test of Time Award in 2017 and the Marco Ramoni best paper award at the 19th International Conference on Artificial Intelligence in Medicine (AIME 2021).