A deep neural spoiler detection model using a genre-aware attention mechanism

Buru Chang, Hyunjae Kim, Raehyun Kim, Deahan Kim, Jaewoo Kang

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    6 Citations (Scopus)

    Abstract

    The fast-growing volume of online activity and user-generated content increases the chances of users being exposed to spoilers. To address this problem, several spoiler detection models have been proposed. However, most of the previous models rely on hand-crafted domain-specific features, which limits the generalizability of the models. In this paper, we propose a new deep neural spoiler detection model that uses a genre-aware attention mechanism. Our model consists of a genre encoder and a sentence encoder. The genre encoder is used to extract a genre feature vector from given genres using a convolutional neural network. The sentence encoder is used to extract sentence feature vectors from a given sentence using a bi-directional gated recurrent unit. We also propose a genre-aware attention layer based on the attention mechanism that utilizes genre information for detecting spoilers which vary by genres. Using a sentence feature, our proposed model determines whether a given sentence is a spoiler. The experimental results on a spoiler dataset show that our proposed model which does not use hand-crafted features outperforms the state-of-the-art spoiler detection baseline models. We also conduct a qualitative analysis on the relations between spoilers and genres, and highlight the results through an attention weight visualization.

    Original languageEnglish
    Title of host publicationAdvances in Knowledge Discovery and Data Mining - 22nd Pacific-Asia Conference, PAKDD 2018, Proceedings
    EditorsDinh Phung, Geoffrey I. Webb, Bao Ho, Vincent S. Tseng, Mohadeseh Ganji, Lida Rashidi
    PublisherSpringer Verlag
    Pages183-195
    Number of pages13
    ISBN (Print)9783319930336
    DOIs
    Publication statusPublished - 2018
    Event22nd Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining, PAKDD 2018 - Melbourne, Australia
    Duration: 2018 Jun 32018 Jun 6

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume10937 LNAI
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Other

    Other22nd Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining, PAKDD 2018
    Country/TerritoryAustralia
    CityMelbourne
    Period18/6/318/6/6

    Bibliographical note

    Publisher Copyright:
    © Springer International Publishing AG, part of Springer Nature 2018.

    Keywords

    • Attention mechanism
    • Classification
    • Deep learning
    • Spoiler alert
    • Spoiler detection

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • General Computer Science

    Fingerprint

    Dive into the research topics of 'A deep neural spoiler detection model using a genre-aware attention mechanism'. Together they form a unique fingerprint.

    Cite this