Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model

Chia Hui Chang; Yu Ching Liao; Ting Yeh

doi:10.1007/978-3-031-20891-1_42

Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model

Chia Hui Chang, Yu Ching Liao, Ting Yeh

Department of Computer Science and Information Engineering

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

Abstract

The problem of finding event announcement pages for any given website is called event source page discovery. In this paper, we show a policy-based deep reinforcement learning (RL) model for the event source page discovery agent. We use two stages to train our agent, pre-training and fine-tuning. In the pre-training phase, the model is trained with limited labeled data, where each episode has a fixed number of steps. In the fine-tuning phase, the agent is trained using unlabeled data and a reward system based on an event source page classifier. The agent learns whether to continue exploring or stop exploring through an adaptive threshold. The proposed agent achieves 74% precision with a 1.28 unit cost (the average number of clicks for each event source page) on the real word data set.

Original language	English
Title of host publication	Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings
Editors	Richard Chbeir, Helen Huang, Fabrizio Silvestri, Yannis Manolopoulos, Yanchun Zhang, Yanchun Zhang
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	597-606
Number of pages	10
ISBN (Print)	9783031208904
DOIs	https://doi.org/10.1007/978-3-031-20891-1_42
State	Published - 2022
Event	23rd International Conference on Web Information Systems Engineering, WISE 2021 - Biarritz, France Duration: 1 Nov 2022 → 3 Nov 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13724 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	23rd International Conference on Web Information Systems Engineering, WISE 2021
Country/Territory	France
City	Biarritz
Period	1/11/22 → 3/11/22

Keywords

Event source page discovery
Multi-task neural model
Reinforcement learning
Web mining

Access to Document

10.1007/978-3-031-20891-1_42

Eventgo: Constructing an Event Search Engine via Event Extraction from Social-Media Posts and Event Source Discovery(2/3)
Chang, C.-H. (PI)
1/08/21 → 31/07/22
Project: Research

Cite this

Chang, C. H., Liao, Y. C., & Yeh, T. (2022). Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model. In R. Chbeir, H. Huang, F. Silvestri, Y. Manolopoulos, Y. Zhang, & Y. Zhang (Eds.), Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings (pp. 597-606). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13724 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-20891-1_42

Chang, Chia Hui ; Liao, Yu Ching ; Yeh, Ting. / Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model. Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings. editor / Richard Chbeir ; Helen Huang ; Fabrizio Silvestri ; Yannis Manolopoulos ; Yanchun Zhang ; Yanchun Zhang. Springer Science and Business Media Deutschland GmbH, 2022. pp. 597-606 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{cf25be36d4214adc970f28dffc610f1a,

title = "Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model",

abstract = "The problem of finding event announcement pages for any given website is called event source page discovery. In this paper, we show a policy-based deep reinforcement learning (RL) model for the event source page discovery agent. We use two stages to train our agent, pre-training and fine-tuning. In the pre-training phase, the model is trained with limited labeled data, where each episode has a fixed number of steps. In the fine-tuning phase, the agent is trained using unlabeled data and a reward system based on an event source page classifier. The agent learns whether to continue exploring or stop exploring through an adaptive threshold. The proposed agent achieves 74% precision with a 1.28 unit cost (the average number of clicks for each event source page) on the real word data set.",

keywords = "Event source page discovery, Multi-task neural model, Reinforcement learning, Web mining",

author = "Chang, {Chia Hui} and Liao, {Yu Ching} and Ting Yeh",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 23rd International Conference on Web Information Systems Engineering, WISE 2021 ; Conference date: 01-11-2022 Through 03-11-2022",

year = "2022",

doi = "10.1007/978-3-031-20891-1_42",

language = "???core.languages.en_GB???",

isbn = "9783031208904",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "597--606",

editor = "Richard Chbeir and Helen Huang and Fabrizio Silvestri and Yannis Manolopoulos and Yanchun Zhang and Yanchun Zhang",

booktitle = "Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings",

}

Chang, CH, Liao, YC & Yeh, T 2022, Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model. in R Chbeir, H Huang, F Silvestri, Y Manolopoulos, Y Zhang & Y Zhang (eds), Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13724 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 597-606, 23rd International Conference on Web Information Systems Engineering, WISE 2021, Biarritz, France, 1/11/22. https://doi.org/10.1007/978-3-031-20891-1_42

Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model. / Chang, Chia Hui; Liao, Yu Ching; Yeh, Ting.
Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings. ed. / Richard Chbeir; Helen Huang; Fabrizio Silvestri; Yannis Manolopoulos; Yanchun Zhang; Yanchun Zhang. Springer Science and Business Media Deutschland GmbH, 2022. p. 597-606 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13724 LNCS).

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

TY - GEN

T1 - Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model

AU - Chang, Chia Hui

AU - Liao, Yu Ching

AU - Yeh, Ting

PY - 2022

Y1 - 2022

N2 - The problem of finding event announcement pages for any given website is called event source page discovery. In this paper, we show a policy-based deep reinforcement learning (RL) model for the event source page discovery agent. We use two stages to train our agent, pre-training and fine-tuning. In the pre-training phase, the model is trained with limited labeled data, where each episode has a fixed number of steps. In the fine-tuning phase, the agent is trained using unlabeled data and a reward system based on an event source page classifier. The agent learns whether to continue exploring or stop exploring through an adaptive threshold. The proposed agent achieves 74% precision with a 1.28 unit cost (the average number of clicks for each event source page) on the real word data set.

AB - The problem of finding event announcement pages for any given website is called event source page discovery. In this paper, we show a policy-based deep reinforcement learning (RL) model for the event source page discovery agent. We use two stages to train our agent, pre-training and fine-tuning. In the pre-training phase, the model is trained with limited labeled data, where each episode has a fixed number of steps. In the fine-tuning phase, the agent is trained using unlabeled data and a reward system based on an event source page classifier. The agent learns whether to continue exploring or stop exploring through an adaptive threshold. The proposed agent achieves 74% precision with a 1.28 unit cost (the average number of clicks for each event source page) on the real word data set.

KW - Event source page discovery

KW - Multi-task neural model

KW - Reinforcement learning

KW - Web mining

UR - http://www.scopus.com/inward/record.url?scp=85142712938&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-20891-1_42

DO - 10.1007/978-3-031-20891-1_42

M3 - 會議論文篇章

AN - SCOPUS:85142712938

SN - 9783031208904

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 597

EP - 606

BT - Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings

A2 - Chbeir, Richard

A2 - Huang, Helen

A2 - Silvestri, Fabrizio

A2 - Manolopoulos, Yannis

A2 - Zhang, Yanchun

PB - Springer Science and Business Media Deutschland GmbH

T2 - 23rd International Conference on Web Information Systems Engineering, WISE 2021

Y2 - 1 November 2022 through 3 November 2022

ER -

Chang CH, Liao YC, Yeh T. Event Source Page Discovery via Policy-Based RL with Multi-task Neural Sequence Model. In Chbeir R, Huang H, Silvestri F, Manolopoulos Y, Zhang Y, Zhang Y, editors, Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 597-606. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-20891-1_42