On Including the user dynamic in learning to rank

Nicola Ferro; Claudio Lucchese; Maria Maistro; Raffaele Perego

doi:10.1145/3077136.3080714

On Including the user dynamic in learning to rank

Nicola Ferro, Claudio Lucchese, Maria Maistro, Raffaele Perego

5 Citations (Scopus)

Abstract

Ranking query results effectively by considering user past behaviour and preferences is a primary concern for IR researchers both in academia and industry. In this context, LtR is widely believed to be the most effective solution to design ranking models that account for user-interaction features that have proved to remarkably impact on IR effectiveness. In this paper, we explore the possibility of integrating the user dynamic directly into the LtR algorithms. Specifically, we model with Markov chains the behaviour of users in scanning a ranked result list and we modify LambdaMart, a state-of-The-Art LtR algorithm, to exploit a new discount loss function calibrated on the proposed Markovian model of user dynamic. We evaluate the performance of the proposed approach on publicly available LtR datasets, finding that the improvements measured over the standard algorithm are statistically significant.

Original language	English
Title of host publication	SIGIR 2017 - Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval
Number of pages	4
Publisher	Association for Computing Machinery, Inc.
Publication date	7 Aug 2017
Pages	1041-1044
ISBN (Electronic)	9781450350228
DOIs	https://doi.org/10.1145/3077136.3080714
Publication status	Published - 7 Aug 2017
Externally published	Yes
Event	40th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2017 - Tokyo, Shinjuku, Japan Duration: 7 Aug 2017 → 11 Aug 2017

Conference

Conference	40th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2017
Country/Territory	Japan
City	Tokyo, Shinjuku
Period	07/08/2017 → 11/08/2017
Sponsor	Association for Computing Machinery Special Interest Group on Information Retrieval (ACM SIGIR)

Keywords

LambdaMart
Learning to rank
User dynamic

Access to Document

10.1145/3077136.3080714

Cite this

On Including the user dynamic in learning to rank. / Ferro, Nicola; Lucchese, Claudio; Maistro, Maria et al.
SIGIR 2017 - Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc., 2017. p. 1041-1044.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Ferro, N, Lucchese, C, Maistro, M & Perego, R 2017, On Including the user dynamic in learning to rank. in SIGIR 2017 - Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc., pp. 1041-1044, 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2017, Tokyo, Shinjuku, Japan, 07/08/2017. https://doi.org/10.1145/3077136.3080714

@inproceedings{8ba492b17ae6446f81ecee093d74f940,

title = "On Including the user dynamic in learning to rank",

abstract = "Ranking query results effectively by considering user past behaviour and preferences is a primary concern for IR researchers both in academia and industry. In this context, LtR is widely believed to be the most effective solution to design ranking models that account for user-interaction features that have proved to remarkably impact on IR effectiveness. In this paper, we explore the possibility of integrating the user dynamic directly into the LtR algorithms. Specifically, we model with Markov chains the behaviour of users in scanning a ranked result list and we modify LambdaMart, a state-of-The-Art LtR algorithm, to exploit a new discount loss function calibrated on the proposed Markovian model of user dynamic. We evaluate the performance of the proposed approach on publicly available LtR datasets, finding that the improvements measured over the standard algorithm are statistically significant.",

keywords = "LambdaMart, Learning to rank, User dynamic",

author = "Nicola Ferro and Claudio Lucchese and Maria Maistro and Raffaele Perego",

year = "2017",

month = aug,

day = "7",

doi = "10.1145/3077136.3080714",

language = "English",

pages = "1041--1044",

booktitle = "SIGIR 2017 - Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval",

publisher = "Association for Computing Machinery, Inc.",

note = "40th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2017 ; Conference date: 07-08-2017 Through 11-08-2017",

}

TY - GEN

T1 - On Including the user dynamic in learning to rank

AU - Ferro, Nicola

AU - Lucchese, Claudio

AU - Maistro, Maria

AU - Perego, Raffaele

PY - 2017/8/7

Y1 - 2017/8/7

N2 - Ranking query results effectively by considering user past behaviour and preferences is a primary concern for IR researchers both in academia and industry. In this context, LtR is widely believed to be the most effective solution to design ranking models that account for user-interaction features that have proved to remarkably impact on IR effectiveness. In this paper, we explore the possibility of integrating the user dynamic directly into the LtR algorithms. Specifically, we model with Markov chains the behaviour of users in scanning a ranked result list and we modify LambdaMart, a state-of-The-Art LtR algorithm, to exploit a new discount loss function calibrated on the proposed Markovian model of user dynamic. We evaluate the performance of the proposed approach on publicly available LtR datasets, finding that the improvements measured over the standard algorithm are statistically significant.

AB - Ranking query results effectively by considering user past behaviour and preferences is a primary concern for IR researchers both in academia and industry. In this context, LtR is widely believed to be the most effective solution to design ranking models that account for user-interaction features that have proved to remarkably impact on IR effectiveness. In this paper, we explore the possibility of integrating the user dynamic directly into the LtR algorithms. Specifically, we model with Markov chains the behaviour of users in scanning a ranked result list and we modify LambdaMart, a state-of-The-Art LtR algorithm, to exploit a new discount loss function calibrated on the proposed Markovian model of user dynamic. We evaluate the performance of the proposed approach on publicly available LtR datasets, finding that the improvements measured over the standard algorithm are statistically significant.

KW - LambdaMart

KW - Learning to rank

KW - User dynamic

UR - http://www.scopus.com/inward/record.url?scp=85029368520&partnerID=8YFLogxK

U2 - 10.1145/3077136.3080714

DO - 10.1145/3077136.3080714

M3 - Article in proceedings

AN - SCOPUS:85029368520

SP - 1041

EP - 1044

BT - SIGIR 2017 - Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval

PB - Association for Computing Machinery, Inc.

T2 - 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2017

Y2 - 7 August 2017 through 11 August 2017

ER -

On Including the user dynamic in learning to rank

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this