ABSTRACT
This chapter introduces the principle learning to rank and its interest into the information retrieval. It discusses the problem of ranking of alternatives, the RankBoost algorithm and the principle of semi-supervised learning. The chapter presents an active learning method of ranking functions of alternatives. Semi-supervised learning is a well-known strategy to label unlabelled data using certain techniques and increase the amount of labelled training data. Learning to rank is a newly popular topic in machine learning. When it is applied to document retrieval, it can be described as the following problem: assume that there is a collection of documents. RankBoost is a supervised learning algorithm of instances designed for ranking problems. It builds a document ranking function by combining a set of ranking features of a set of document pairs. The RankBoostSSA algorithm can implicitly select characteristics. This property can be an advantage for this algorithm to the supervised method.
