Abstract
Learning-to-rank has been intensively studied and has shown significantly increasing values in a wide range of domains, such as web search, recommender systems, dialogue systems, machine translation, and even computational biology, to name a few. In light of recent advances in neural networks, there has been a strong and continuing interest in exploring how to deploy popular techniques, such as reinforcement learning and adversarial learning, to solve ranking problems. However, armed with the aforesaid popular techniques, most studies tend to show how effective a new method is. A comprehensive comparison between techniques and an in-depth analysis of their deficiencies are somehow overlooked. This paper is motivated by the observation that recent ranking methods based on either reinforcement learning or adversarial learning boil down to policy-gradient-based optimization. Based on the widely used benchmark collections with complete information (where relevance labels are known for all items), such as MSLRWEB30K and Yahoo-Set1, we thoroughly investigate the extent to which policy-gradient-based ranking methods are effective. On one hand, we analytically identify the pitfalls of policy-gradient-based ranking. On the other hand, we experimentally compare a wide range of representative methods. The experimental results echo our analysis and show that policy-gradient-based ranking methods are, by a large margin, inferior to many conventional ranking methods. Regardless of whether we use reinforcement learning or adversarial learning, the failures are largely attributable to the gradient estimation based on sampled rankings, which significantly diverge from ideal rankings. In particular, the larger the number of documents per query and the more fine-grained the ground-truth labels, the greater the impact policy-gradient-based ranking suffers. Careful examination of this weakness is highly recommended for developing enhanced methods based on policy gradient.
Highlights
IntroductionLearning-to-rank has long been studied, with applications spanning across many fields, such as web search, dialogue systems, and computational biology [1]
Introduction published maps and institutional affilLearning-to-rank has long been studied, with applications spanning across many fields, such as web search, dialogue systems, and computational biology [1]
We focus on datasets consisting of feature vectors and refer the reader to the study in [61] for a comprehensive overview of bidirectional encoder representations from transformers (BERT)-based learning-to-rank
Summary
Learning-to-rank has long been studied, with applications spanning across many fields, such as web search, dialogue systems, and computational biology [1]. Each query is associated with a set of documents to be ranked, for which the standard relevance labels are included. The desired scoring model (or function) assigns a score to each document, a ranked list of documents can be obtained by sorting the documents in descending order of scores. The document with the highest score is assigned a rank of 1. The rank position of a document represents its relevance with respect to the query. The metrics, such as average precision (AP) and normalized discounted cumulative gain (nDCG) [2], are adopted to measure the performance
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
