Boosting learning to rank with user dynamics and continuation methods
- PDF / 1,632,351 Bytes
- 27 Pages / 439.37 x 666.142 pts Page_size
- 87 Downloads / 224 Views
Boosting learning to rank with user dynamics and continuation methods Nicola Ferro1 · Claudio Lucchese2 · Maria Maistro1,3 · Raffaele Perego4 Received: 6 March 2019 / Accepted: 13 October 2019 © Springer Nature B.V. 2019
Abstract Learning to rank (LtR) techniques leverage assessed samples of query-document relevance to learn effective ranking functions able to exploit the noisy signals hidden in the features used to represent queries and documents. In this paper we explore how to enhance the state-of-the-art LambdaMart LtR algorithm by integrating in the training process an explicit knowledge of the underlying user-interaction model and the possibility of targeting different objective functions that can effectively drive the algorithm towards promising areas of the search space. We enrich the iterative process followed by the learning algorithm in two ways: (1) by considering complex query-based user dynamics instead than simply discounting the gain by the rank position; (2) by designing a learning path across different loss functions that can capture different signals in the training data. Our extensive experiments, conducted on publicly available datasets, show that the proposed solution permits to improve various ranking quality measures by statistically significant margins. Keywords Learning to rank · User dynamics · Continuation methods
* Maria Maistro [email protected]; [email protected] Nicola Ferro [email protected] Claudio Lucchese [email protected] Raffaele Perego [email protected] 1
Department of Information Engineering, University of Padua, Via Gradenigo 6/b, 35131 Padua, PD, Italy
2
Department of Environmental Sciences, Informatics and Statistics, Ca’ Foscari University of Venice, Via Torino 155, 30172 Mestre, VE, Italy
3
Department of Computer Science, University of Copenhagen, Universitetsparken 1, 2100 Copenhagen, Denmark
4
Istituto di Scienza e Tecnologie dell’Informazione A. Faedo (ISTI), National Research Council (CNR), Area di Ricerca di Pisa, Via G. Moruzzi 1, 56124 Pisa, PI, Italy
13
Vol.:(0123456789)
Information Retrieval Journal
1 Introduction Information retrieval (IR) systems are nowadays challenged with increasingly complex search tasks where information about how users interact with the system plays a central role to adapt them to the needs and interests of users. A lot of research efforts focused on enhancing user engagement and retrieval effectiveness by exploiting information about user-system interactions available, for example, from the query logs of Web search engines (Silvestri 2009; Lucchese et al. 2013; Mehrotra et al. 2018). The ability to interpret and learn from the complex and noisy traces of user-system interactions is fundamental for IR advance. The number of clicks on a given query-result pair, the click-through rate (CTR), and the dwell time, are examples of actionable information to improve various aspects of IR systems (Chuklin et al. 2015). In the context of learning to rank (LtR) (Liu 2009), user actions recorded in query logs
Data Loading...
