HotelRec: a Novel Very Large-Scale Hotel Recommendation Dataset

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4917–4923 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4917 HotelRec: a Novel Very Large-Scale Hotel Recommendation Dataset Diego Antognini, Boi Faltings Artificial Intelligence Laboratory ´Ecole Polytechnique F´ed´erale de Lausanne Lausanne, Switzerland firstname.lastname@epfl.ch Abstract Today, recommender systems are an inevitable part of everyone’s daily digital routine and are present on most internet platforms. State-of-the-art deep learning-based models require a large number of data to achieve their best performance. Many datasets fulfilling this criterion have been proposed for multiple domains, such as Amazon products, restaurants, or beers. However, works and datasets in the hotel domain are limited: the largest hotel review dataset is below the million samples. Additionally, the hotel domain suffers from a higher data sparsity than traditional recommendation datasets and therefore, traditional collaborative-filtering approaches cannot be applied to such data. In this paper, we propose HotelRec, a very large-scale hotel recommendation dataset, based on TripAdvisor, containing 50 million reviews. To the best of our knowledge, HotelRec is the largest publicly available dataset in the hotel domain (50M versus 0.9M) and additionally, the largest recommendation dataset in a single domain and with textual reviews (50M versus 22M). We release HotelRec for further research: https://github.com/Diego999/HotelRec. Keywords: reviews, recommender systems, text mining, sentiment analysis 1. Introduction The increasing flood of information on the web creates a need for selecting content according to the end user’s pref- erences. Today, recommender systems are deployed on most internet platforms and play an important role in ev- erybody’s daily digital routine, including e-commerce web- sites, social networks, music streaming, or hotel booking. Recommender systems have been investigated over more than thirty years (Bobadilla et al., 2013). Over the years, many models and datasets in different domains and vari- ous sizes have been developed: movies (Harper and Kon- stan, 2016), Amazon products (McAuley et al., 2015; He and McAuley, 2016), or music (Celma, 2010). With the tremendous success of large deep learning-based rec- ommender systems, in better capturing user-item interac- tions, the recommendation quality has been significantly improved (Covington et al., 2016). However, the increase in recommendation performance with deep learning-based models comes at the cost of large datasets. Most recent state-of-the-art models, such as (Wang et al., 2019), (Liang et al., 2018), or (He et al., 2017) necessitate large datasets (i.e., millions) to achieve high performance. In the hotel domain, only a few works have studied hotel recommendation, such as (Wang et al., 2011) or (Zhang et al., 2015). Additionally, to the best of our knowledge, the largest publicly available hotel review dataset contains 870k samples (Li et al., 2016). Unlike commonly used recommendation datasets, the hotel domain suffers from higher data sparsity and therefore, traditional collaborative- filtering approaches cannot be applied (Zhang et al., 2015; Khaleghi et al., 2018; Musat and Faltings, 2015). Further- more, rating a hotel is different than traditional products, because the whole experience lasts longer, and there are more facets to review (Khaleghi …

Content truncated to 500 words