Foundations of Machine Learning for IR
Abstract: Passage retrieval, question answering, and language generation – those are some of the tasks developed for MS MARCO, one of the very few leaderboard-based benchmarks that have a heavy focus on IR tasks. The large-scale nature of the data naturally leads to machine learning (ML) based solutions. In this lecture, we will take a closer look at some of the best performing ML approaches which are based on learning to rank and neural IR, the latter often heavily inspired by the most recent advances in NLP. We will conclude with a critical discussion of the recent neural net wave and its applicability to a wide(r) range of IR tasks.
Short Bio: Claudia Hauff is an Associate Professor at the Web Information Systems group, Delft University of Technology (TU Delft) and a computer scientist by training. She received her PhD in 2010 from the University of Twente. In the past, she has worked on a variety of topics in the fields of information retrieval & data science, including query performance prediction, social search, learning to search and information retrieval for specific user groups. Together with a number of PhD students she currently focuses on the areas of collaborative search, complex search, and conversational search.