Stefanos Angelidis

I am a Machine Learning and Natural Language Processing researcher and engineer, with years of experience developing novel deep learning solutions in academic and industrial settings. I am currently a Senior Mechine Learning Engineer and Team Lead in Taboola (part of Deeplab's passionate team of Taboolars), helping build a state-of-the-art recommender system, which serves many billions requests per day.

Before this, I served as a Postdoctoral Research Fellow in the EdinburghNLP group (part of the Institute for Language, Cognition and Computation, School of Informatics, University of Edinburgh). During my time there, I was fortunate to be advised by and collaborate with Prof. Mirella Lapata, and had years-long synergies with other members of her team. I've also worked with Megagon Labs and Amazon CoreML (Barcelona).

My academic research focused on deep learning methods for understanding long-form text with applications on summarization, opinion mining and question answering. I became particularly interested in applying novel solutions to language understanding problems where task-specific supervision is scarce. While many of my core interests still lie in NLP, I am constantly trying to broaden my expertise in other ML applications, including recommender systems, information retrieval and multi-modal ML.

I succesfully completed my PhD in April 2019, under the supervision of Prof. Mirella Latapa and Prof. Charles Sutton (Thesis: "Weakly Supervised Sentiment Analysis and Opinion Extraction"). Before my PhD, I obtained a MSc in Artificial Intelligence with distinction from the University of Edinburgh. My MSc Thesis was supervised by Prof. Victor Lavrenko. I received my undergraduate degree on Informatics and Telecommunucations from the National and Kapodistrian University of Athens.

Published and Presented Work

Extractive Opinion Summarization in Quantized Transformer Spaces.
Stefanos Angelidis, Reinald Kim Amplayo, Yoshihiko Suhara, Xiaolan Wang, Mirella Lapata. In Transactions of the Association for Computational Linguistics (TACL)
[ code & data ]

Unsupervised Opinion Summarization with Content Planning.
Reinald Kim Amplayo, Stefanos Angelidis, Mirella Lapata. In AAAI 2021.
[ code ]

OpinionDigest: A Simple Framework for Opinion Summarization.
Yoshihiko Suhara, Xiaolan Wang, Stefanos Angelidis, Wang-Chiew Tan. 2020. In ACL 2020.
[ bibtex ] [ code ]

ExtremeReader: An interactive explorer for customizable and explainable review summarization.
Xiaolan Wang, Yoshihiko Suhara, Natalie Nuno, Yuliang Li, Jinfeng Li, Nofar Carmeli, Stefanos Angelidis, Eser Kandogan and Wang-Chiew Tan. 2020. In WWW 2020 (demo).
[ bibtex ] [ video ]

Book QA: Stories of Challenges and Opportunities.
Stefanos Angelidis, Lea Frermann, Diego Marcheggiani, Roi Blanco, Lluís Màrquez. 2019. In 2nd Workshop on Machine Reading for Question Answering.
[ bibtex ] [ data ] [ poster ]

Summarizing Opinions: Aspect Extraction Meets Sentiment Prediction and They Are Both Weakly Supervised.
Stefanos Angelidis, Mirella Lapata. 2018. In EMNLP 2018.
[ bibtex ] [ Appendix ] [ data + code ] [ slides ]

Multiple Instance Learning Networks for Fine-grained Sentiment Analysis.
Stefanos Angelidis, Mirella Lapata. In Transactions of the Association for Computational Linguistics (TACL) vol. 6, 2018.
[ bibtex ] [ data ] [ code ] [ slides ] [ video ]

Document Embeddings with Context Sampling.
Stefanos Angelidis, Victor Lavrenko, Mirella Lapata. Presented Poster. Language @ Edinburgh, June 2016.

Learning to Segment Web-Pages Using Repeat Analysis.
Stefanos Angelidis. MSc Thesis. Supervised by Victor Lavrenko.