A Topic Modeling Approach to Ranking
This work addresses preference prediction for users in ranking tasks, but it is incremental as it builds on existing topic modeling techniques.
The authors tackled the problem of predicting preferences in pairwise comparisons by developing a new generative model that accounts for multiple shared latent rankings and inconsistent user behavior, and they demonstrated that this approach is empirically competitive with state-of-the-art methods on semi-synthetic and real-world datasets.
We propose a topic modeling approach to the prediction of preferences in pairwise comparisons. We develop a new generative model for pairwise comparisons that accounts for multiple shared latent rankings that are prevalent in a population of users. This new model also captures inconsistent user behavior in a natural way. We show how the estimation of latent rankings in the new generative model can be formally reduced to the estimation of topics in a statistically equivalent topic modeling problem. We leverage recent advances in the topic modeling literature to develop an algorithm that can learn shared latent rankings with provable consistency as well as sample and computational complexity guarantees. We demonstrate that the new approach is empirically competitive with the current state-of-the-art approaches in predicting preferences on some semi-synthetic and real world datasets.