Java: Data Science Made Easy

Learning to rank for information retrieval

Learning to rank is a family of algorithms that deal with ordering data. This family is a part of supervised machine learning; to order the data, we need to know which items are more important and need to be shown first.

Learning to rank is often used in the context of building search engines; based on some relevance evaluations, we build a model that tries to rank relevant items higher than nonrelevant ones. In the unsupervised ranking case, such as cosine on TF-IDF weights, we typically have only one feature, by which we order the documents. However, there could be a lot more features, which we may want to include in the model and let it combine them in the best possible way.

There are several ...