Learning Determinantal Point Processes

TL;DR

This paper introduces a convex, efficient method for learning determinantal point processes (DPPs) from data, enabling diverse subset selection, and demonstrates its effectiveness in extractive summarization with state-of-the-art results.

Contribution

It proposes a natural feature-based parameterization of conditional DPPs that facilitates convex learning and applies it to multi-document summarization.

Findings

Achieved state-of-the-art results on DUC 2003/04 datasets.

Established a convex, efficient learning framework for DPPs.

Demonstrated effective balancing of relevance and diversity in summarization.

Abstract

Determinantal point processes (DPPs), which arise in random matrix theory and quantum physics, are natural models for subset selection problems where diversity is preferred. Among many remarkable properties, DPPs offer tractable algorithms for exact inference, including computing marginal probabilities and sampling; however, an important open question has been how to learn a DPP from labeled training data. In this paper we propose a natural feature-based parameterization of conditional DPPs, and show how it leads to a convex and efficient learning formulation. We analyze the relationship between our model and binary Markov random fields with repulsive potentials, which are qualitatively similar but computationally intractable. Finally, we apply our approach to the task of extractive summarization, where the goal is to choose a small subset of sentences conveying the most important information from a set of documents. In this task there is a fundamental tradeoff between sentences that are highly relevant to the collection as a whole, and sentences that are diverse and not repetitive. Our parameterization allows us to naturally balance these two characteristics. We evaluate our system on data from the DUC 2003/04 multi-document summarization task, achieving state-of-the-art results.