Paper on Interpretable Sorting of Multiple Attributes published at TVCG

Consider searching for cheap hotel at a good location on a travel website. You can sort hotels by price, but the locations can either be near or far from your desired location. Conversely, you can sort by distance, but the prices will vary wildly. Why can’t we have the sort by both price and distance?

Now you can! With Imma Sort.

We introduce Interpretable Monotonic Multi-Attribute Sorting, a technique to support sorting by multiple attributes simultaneously. We identified that people appreciate sorted lists because the monotonicity in trends allow easy predictability of what values will come next or later, which allows us to anticipate when to stop our search. However, We further realize that this predictability does not need to be exact and can be relaxed. Therefore, we leveraged the idea of approximate monotonicity, where decreasing the monotonocity of one attribute (e.g., price) can enable the increase of monotonocity of another attribute (e.g., distance). Mathematically, we implemented this using Ranking Principal Curves, and added a tunable parameter to weighted which attribute to prioritize for smoothness.

With our technique, Imma sort food meals by taste and health, and Imma sort movies by rating and popularity.

In modeling and user studies, we show that Imma Sort is intuitive and usable, and improves the predictive interpretability and satisfaction of sorting by >2 attributes.

We will present our paper at IEEE VIS in October later this year. Congratulations to team members Lyu Yan, Fan Gao, and I-Shuen Wu!

IEEE VIS presentation:

Yan Lyu, Fan Gao, I-Shuen Wu, and Brian Y. Lim. 2020. Imma Sort by two or more attributes with Interpretable Monotonic Multi-Attribute Sorting. Accepted to IEEE Transactions on Visualization and Computer Graphics (TVCG).