From amateurs to connoisseurs: modeling the evolution of user expertise through online reviews
By Julian McAuley, Jure Leskovec. 2013
700 citations recorded by the source corpus.
Recommending products to consumers means not only understanding their tastes, but also understanding their level of experience. For example, it would be a mistake to recommend the iconic film Seven Samurai simply because a user enjoys other action movies; rather, we might conclude that they will eventually enjoy it---once they are ready. The same is true for beers, wines, gourmet foods---or any products where users have acquired tastes: the `best' products may not be the most 'accessible'. Thus our goal in this paper is to recommend products that a user will enjoy now, while acknowledging that their tastes may have changed over time, and may change again in the future. We model how tastes change due to the very act of consuming more products---in other words, as users become more experienced. We develop a latent factor recommendation system that explicitly accounts for each user's level of experience. We find that such a model not only leads to better recommendations, but also allows us to study the role of user experience and expertise on a novel dataset of fifteen million beer, wine, food, and movie reviews.
Subfields: Artificial Intelligence, Information Systems, Tourism, Leisure and Hospitality Management, Sentiment & Opinion Analysis, Recommender Systems
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Sebastian Ruder
- Jianfeng Gao
- Noah A. Smith
- Luke S. Zettlemoyer
- Yoshua Bengio
- Philip S. Yu
- Jure Leskovec
In Information Systems
- Srijan Kumar
- Philip S. Yu
- Jure Leskovec
- Julian McAuley
- Ruining He
- Anders Søgaard
- Xiaojun Wan
- Ke Wang
In Tourism, Leisure and Hospitality Management
In Sentiment & Opinion Analysis
In Recommender Systems
Citations in the corpus
- Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering (2016, methodology): “This procedure resembles the one used in [25], though the problem setting and data are different.”
- Square One Bias in NLP: Towards a Multi-Dimensional Exploration of the Research Manifold (2022, background): “Annotation biases can also stem from the characteristics of the annotators, including their domain experience (McAuley and Leskovec, 2013), demographics (Jørgensen and Søgaard, 2021), or educational level (Al Kuwatly et al.”
- Comparing the Predictive Capability of Social and Interest Affinity for Recommendations (2014, background): “Is the performance independent of users selectiveness or items likeability? Only few studies hint at the relation between user selectiveness [34]”
- Hidden factors and hidden topics: understanding rating dimensions with review text (2013, methodology): “We also consider 6 million beer and wine reviews previously studied in [19, 21], though in addition we include pub data from ratebeer.”
- REV2: Fraudulent User Prediction in Rating Platforms (2018, background): “• Amazon is a user-to-product rating network [20].”
- Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters (2020, methodology): “We use the Yelp review dataset [29] and Amazon review dataset [26] to study the fraudster camouflage and GNNbased fraud detection problem.”
- No country for old members: user lifecycle and linguistic change in online communities (2013, background): “There is extensive research covering the evolution of online communities [21, 28], the evolution of community members [2, 10, 16, 29, 32], and their participation patterns [1, 3, 26, 30, 31, 48] (see Section 6 for a discussion).”
- Diversity-Promoting GAN: A Cross-Entropy Based Generative Adversarial Network for Diversified Text Generation (2018, methodology): “Amazon Review Generation Dataset (Amazon): This dataset is provided by McAuley and Leskovec (2013).”
- Pick and Choose: A GNN-based Imbalanced Learning Approach for Fraud Detection (2021, methodology): “The Amazon dataset[28] includes product reviews under the Musical Instrument category.”
- SentiGAN: Generating Sentimental Texts via Mixture Adversarial Networks (2018, methodology): “We use the data scraped from Beer-Advocate [Mcauley and Leskovec, 2013].”
- Unpaired Sentiment-to-Sentiment Translation: A Cycled Reinforcement Learning Approach (2018, methodology): “Amazon Food Review Dataset (Amazon): This dataset is provided by McAuley and Leskovec (2013).”
- Artificial intelligence in recommender systems (2020, background): “In a similar case, user expertise was evaluated and the evolution of user experience was tracked through online reviews, suggesting that similar users with an equivalent level of experience are likely to respond similarly to the same product [186].”
- Winning Arguments: Interaction Dynamics and Persuasion Strategies in Good-faith Online Discussions (2016, background): “It is worth noting that this lack of apparent improvement might be explained by a gradual development of a “taste” for original posts that are harder to address [30].”