Research Presentation by Mizuki Hisatomi at ACM UIST 2025

2025.9.30

Her research concerns a method for automatically grouping web browser tabs that are related to the same task, in order to solve the problem of finding the desired tab becoming difficult when many tabs (e.g., several dozens) are open simultaneously in a web browser.

Existing research includes methods such as creating a classifier using machine learning that, given a sequence of tab transitions, determines where a transition is to another tab within the same task and where it is to a tab belonging to a different task. The proposed method applies a community detection algorithm to a directed graph representing the transitions between tabs, thereby extracting the set of tabs corresponding to each task as a community. A community detection algorithm is one that, given a graph, partitions the set of all vertices into groups of vertices that have dense links among themselves.

Compared to existing machine learning-based methods, our proposed method has the advantages that it does not require prior training and does not require providing personal information, such as the pages accessed, as features to a machine learning algorithm.

This conference was held in Haeundae, a beachside town in Busan, South Korea.
When I arrived, there was some kind of event happening on the beach.

The hotel on the left in the photo below was the venue.

Here are scenes from the conference opening.
Next to the main screen, there was a real-time transcript of the speaker’s remarks displayed as subtitles.

This is the demo session venue. The photo below shows a demonstration by a research group from Nagoya City University that creates the illusion that one’s own skin is stretching, which I’ve seen introduced on Japanese TV before.