Learning to Rank

Table of Contents

Short Summary: Various ranking system models.

Problem: given a query \(q\) and a set of documents \(D = \{d_1, \dots, d_n\}\), rank the documents based on relevance. Put it simply, ranking is a sorting problem, where the key is to find the "key" according to which to sort. The key is typically a Machine Learning (ML) model.

Unfortunately, rank information is available only after sorting, and sorting is non differentiable.

Pointwise Methods

Pairwise Methods

Listwise Methods