On the Role of Reviewer Expertise in Temporal Review Helpfulness Prediction

Findings of the Association for Computational Linguistics: EACL 2023, pages 1684–1692 May 2-6, 2023 ©2023 Association for Computational Linguistics On the Role of Reviewer Expertise in Temporal Review Helpfulness Prediction Mir Tafseer Nayeem University of Alberta mnayeem@ualberta.ca Davood Rafiei University of Alberta drafiei@ualberta.ca Abstract Helpful reviews have been essential for the success of e-commerce services, as they help customers make quick purchase decisions and benefit the merchants in their sales. While many reviews are informative, others provide little value and may contain spam, excessive appraisal, or unexpected biases. With the large volume of reviews and their uneven quality, the problem of detecting helpful reviews has drawn much attention lately. Existing meth- ods for identifying helpful reviews primarily focus on review text and ignore the two key fac- tors of (1) who post the reviews and (2) when the reviews are posted. Moreover, the helpful- ness votes suffer from scarcity for less popular products and recently submitted (a.k.a., cold- start) reviews. To address these challenges, we introduce a dataset and develop a model that in- tegrates the reviewer’s expertise, derived from the past review history of the reviewers, and the temporal dynamics of the reviews to automat- ically assess review helpfulness. We conduct experiments on our dataset to demonstrate the effectiveness of incorporating these factors and report improved results compared to several well-established baselines. 1 Introduction Many customers rely on online reviews from non- professionals, on daily basis, to decide what prod- ucts to buy (e.g., Amazon), what hotels to stay at (e.g., TripAdvisor), what restaurants to eat (e.g., Yelp) and even what books to read (e.g., Goodreads). A recent survey of Bizrate Insights re- ward members found that approximately 98% of on- line shoppers research a vendor via online reviews before making a purchase decision (Kats, 2018). Since the reviews are expected to describe the ac- tual experiences and opinions of users, they can provide a reliable source of reference, improving other customers’ confidence, comfort, and the over- all shopping experience (Foo et al., 2017; Gamzu Best view in town "What can I say .. this is my best place in town. Average food, but the view pays the price. Breathtaking London View, lovely staff, Love to ......." COVID restricted "The room was clean and comfortable. We were looking forward to the breakfast buffet, but due to COVID, it wasn’t available. We didn't dine in for other meals ......." (a) (b) (c) HORRIBLE Service "Terrible food! Overpriced, Cold, and flavorless. Shocking Service!! Undoubtedly the WORST place I have ever been! Call +1 437 *** **** OR visit this restaurant *****..." 0 Figure 1: A snapshot of three reviews with the review- ers’ history information: Review a has accumulated more helpful votes but is posted almost two years before Review b; on the other hand, Review b (a.k.a., cold-start review) contains time-sensitive information, describing the current conditions and Review c is likely a spam re- view. Photos of the reviewers are replaced with avatars for privacy reasons. et al., 2021). However, …

Content truncated to 500 words