Paweł Guzewicz: Guiding Interactive Data Analysis with Context-Sensitive Recommendations

Paweł Guzewicz, PhD student in our team, will do a seminar on November 18 at 9:15am in Thomas Flowers room in Turing building. He will present a series of works by Tova Milo (see the references).

Guiding Interactive Data Analysis with Context-Sensitive Recommendations

Modern Interactive Data Analysis systems offer rich, practical and user-friendly interfaces for performing data exploration, mining and visualization tasks. In spite of that, non-expert users, as well as trained data analysts, still face many challenges examining datasets. In particular, domain knowledge and understanding of the context are essential for meaningful analysis outcomes. Throughout the talk we present various approaches to address the need to support a user in their interactive exploration by guiding the data analysis with context-sensitive recommendations.
We focus on a series of research conducted by Tova Milo et al. that led to a development of REACT system, along with studies on Reinforcement Learning-driven Exploratory Data Analysis, and choosing the interestingness measure based on assessing the interestingness of Data Analysis actions. REACT is a recommender system that generates personalized next-action suggestions to the user. The system is trained on previous analytical sessions with a focus on generalizing relevant candidate actions from the session logs to abstract actions, to be further adapted to the user context. Going beyond, the authors put effort in exploring the area of entirely autonomous solutions for Exploratory Data Analysis by adapting Reinforcement Learning process using an agent to mimic a data analyst performing analytical actions. Finally, we demonstrate an approach for choosing the most adequate interestingness measure with regard to user’s interest at the given step of an interactive analysis session.

[1] Tova Milo and Amit Somech. 2016. REACT: Context-Sensitive Recommendations for Data Analysis. In Proceedings of the 2016 International Conference on Management of Data (SIGMOD ’16). ACM, New York, NY, USA, 2137-2140. DOI:

[2] Tova Milo and Amit Somech. 2018. Next-Step Suggestions for Modern Interactive Data Analysis Platforms. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD ’18). ACM, New York, NY, USA, 576-585. DOI:

[3] Tova Milo and Amit Somech. 2018. Deep Reinforcement-Learning Framework for Exploratory Data Analysis. In Proceedings of the First International Workshop on Exploiting Artificial Intelligence Techniques for Data Management (aiDM’18). ACM, New York, NY, USA, Article 4, 4 pages. DOI:

[4] Somech, Amit, Tova Milo and Chai Ozeri. “Predicting “What is Interesting” by Mining Interactive-Data-Analysis Session Logs.” EDBT (2019).

Comments are closed.