Representation Learning for Users' Web Browsing Sequences

Yukihiro TAGAMI; Hayato KOBAYASHI; Shingo ONO; Akira TAJIMA

doi:10.1587/transinf.2017EDP7335

Representation Learning for Users' Web Browsing Sequences

Yukihiro TAGAMI, Hayato KOBAYASHI, Shingo ONO, Akira TAJIMA

Full Text Views

0

Cite this

Summary :

Modeling user activities on the Web is a key problem for various Web services, such as news article recommendation and ad click prediction. In our work-in-progress paper[1], we introduced an approach that summarizes each sequence of user Web page visits using Paragraph Vector[3], considering users and URLs as paragraphs and words, respectively. The learned user representations are used among the user-related prediction tasks in common. In this paper, on the basis of analysis of our Web page visit data, we propose Backward PV-DM, which is a modified version of Paragraph Vector. We show experimental results on two ad-related data sets based on logs from Web services of Yahoo! JAPAN. Our proposed method achieved better results than those of existing vector models.

Publication: IEICE TRANSACTIONS on Information Vol.E101-D No.7 pp.1870-1879

Publication Date: 2018/07/01

Publicized: 2018/04/20

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2017EDP7335

Type of Manuscript: PAPER

Category: Artificial Intelligence, Data Mining

Authors

Yukihiro TAGAMI
  Yahoo Japan Corporation
Hayato KOBAYASHI
  Yahoo Japan Corporation
Shingo ONO
  Yahoo Japan Corporation
Akira TAJIMA
  Yahoo Japan Corporation

Keyword

online advertising, Web browsing behavior, Paragraph Vector, representation learning

Cite this

Copy

Yukihiro TAGAMI, Hayato KOBAYASHI, Shingo ONO, Akira TAJIMA, "Representation Learning for Users' Web Browsing Sequences" in IEICE TRANSACTIONS on Information, vol. E101-D, no. 7, pp. 1870-1879, July 2018, doi: 10.1587/transinf.2017EDP7335.
Abstract: Modeling user activities on the Web is a key problem for various Web services, such as news article recommendation and ad click prediction. In our work-in-progress paper[1], we introduced an approach that summarizes each sequence of user Web page visits using Paragraph Vector[3], considering users and URLs as paragraphs and words, respectively. The learned user representations are used among the user-related prediction tasks in common. In this paper, on the basis of analysis of our Web page visit data, we propose Backward PV-DM, which is a modified version of Paragraph Vector. We show experimental results on two ad-related data sets based on logs from Web services of Yahoo! JAPAN. Our proposed method achieved better results than those of existing vector models.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2017EDP7335/_p

Copy

@ARTICLE{e101-d_7_1870,
author={Yukihiro TAGAMI, Hayato KOBAYASHI, Shingo ONO, Akira TAJIMA, },
journal={IEICE TRANSACTIONS on Information},
title={Representation Learning for Users' Web Browsing Sequences},
year={2018},
volume={E101-D},
number={7},
pages={1870-1879},
abstract={Modeling user activities on the Web is a key problem for various Web services, such as news article recommendation and ad click prediction. In our work-in-progress paper[1], we introduced an approach that summarizes each sequence of user Web page visits using Paragraph Vector[3], considering users and URLs as paragraphs and words, respectively. The learned user representations are used among the user-related prediction tasks in common. In this paper, on the basis of analysis of our Web page visit data, we propose Backward PV-DM, which is a modified version of Paragraph Vector. We show experimental results on two ad-related data sets based on logs from Web services of Yahoo! JAPAN. Our proposed method achieved better results than those of existing vector models.},
keywords={},
doi={10.1587/transinf.2017EDP7335},
ISSN={1745-1361},
month={July},}

Copy

TY - JOUR
TI - Representation Learning for Users' Web Browsing Sequences
T2 - IEICE TRANSACTIONS on Information
SP - 1870
EP - 1879
AU - Yukihiro TAGAMI
AU - Hayato KOBAYASHI
AU - Shingo ONO
AU - Akira TAJIMA
PY - 2018
DO - 10.1587/transinf.2017EDP7335
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E101-D
IS - 7
JA - IEICE TRANSACTIONS on Information
Y1 - July 2018
AB - Modeling user activities on the Web is a key problem for various Web services, such as news article recommendation and ad click prediction. In our work-in-progress paper[1], we introduced an approach that summarizes each sequence of user Web page visits using Paragraph Vector[3], considering users and URLs as paragraphs and words, respectively. The learned user representations are used among the user-related prediction tasks in common. In this paper, on the basis of analysis of our Web page visit data, we propose Backward PV-DM, which is a modified version of Paragraph Vector. We show experimental results on two ad-related data sets based on logs from Web services of Yahoo! JAPAN. Our proposed method achieved better results than those of existing vector models.
ER -