Anzeige
Erschienen in:
2019 | OriginalPaper | Buchkapitel
Fusion Convolutional Attention Network for Opinion Spam Detection
verfasst von : Jiacheng Li, Qianwen Ma, Chunyuan Yuan, Wei Zhou, Jizhong Han, Songlin Hu
Erschienen in: Neural Information Processing
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
Opinion spam detection is a critical task in opinion mining. Current researches mainly focus on manually designing the text or discrete user features, or concatenating the review text and user features as the representation of the review. However, these methods ignore the impact of user’s preferences on review texts. Because of their different purposes, spammers and non-spammers usually show different preferences. These user-level differences are hard to be captured from the single review level by previous methods.
In this paper, we propose a novel Fusion Convolutional Attention Network (FCAN) to embed the user-level information into a continuous vector space, the representations of which capture essential clues such as user profiles or preferences. Such user representation, in turn, facilitates learning better user-aware textual representation at word and sentence level. Experimental results on four real-world datasets from different platforms demonstrate that our method significantly outperforms the state-of-the-art methods and can be easily expanded to different datasets and platforms.