📊 Datasets — Awesome Recommender Systems

Movie-rating datasets (100K–25M ratings) — the standard recommender-systems benchmark.

Yelp is a platform that contains user-generated reviews and ratings of businesses, which is used to evaluate and improve sentiment-aware recommendation systems.

Amazon Customer Reviews (a.k.a. Product Reviews) is one of Amazons iconic products. In a period of over two decades since the first review in 1995, millions of Amazon customers have contributed over a hundred million reviews to express opinions and describe their experiences regarding products on the Amazon.com website. This makes Amazon Customer Reviews a rich source of information for academic researchers in the fields of Natural Language Processing (NLP), Information Retrieval (IR), and Machine Learning (ML), amongst others. Accordingly, we are releasing this data to further research in multiple disciplines related to understanding customer product experiences. Specifically, this dataset was constructed to represent a sample of customer evaluations and opinions, variation in the perception of a product across geographical regions, and promotional intent or bias in reviews. Over 130+ million customer reviews are available to researchers as part of this release. The data is available in TSV files in the amazon-reviews-pds S3 bucket in AWS US East Region. Each line in the data files corresponds to an individual review (tab delimited, with no quote and escape characters). Each Dataset contains the following columns: - marketplace: 2 letter country code of the marketplace where the review was written. - customer_id: Random identifier that can be used to aggregate reviews written by a single author. - review_id: The unique ID of the review. - product_id: The unique Product ID the review pertains to. In the multilingual dataset the reviews for the same product in different countries can be grouped by the same product_id. - product_parent: Random identifier that can be used to aggregate reviews for the same product. - product_title: Title of the product. - product_category: Broad product category that can be used to group reviews (also used to group the dataset into coherent parts). - star_rating: The 1-5 star rating of the review. - helpful_votes: Number of helpful votes. - total_votes: Number of total votes the review received. - vine: Review was written as part of the Vine program. - verified_purchase: The review is on a verified purchase. - review_headline: The title of the review. - review_body: The review text. - review_date: The date the review was written.

The 'Netflix' dataset is a benchmark containing user ratings for movies, used to evaluate the performance of recommendation systems, particularly in terms of prediction accuracy and diversity.

The 'Amazon-Book' dataset is a benchmark used to evaluate recommender systems, containing user-item interactions specifically related to books.

Last.fm is a dataset used to evaluate recommender systems, containing user listening histories and music preferences to assess the effectiveness of personalized content delivery and recommendation diversity.

The MIND dataset is a benchmark that contains user interaction data for news articles and is used to evaluate the performance of recommender systems, particularly in the context of collaborative filtering.

The 'Taobao' dataset is used to evaluate lifelong user interest modeling in recommender systems, containing user behavior sequences for display advertising.

The 'MovieLens-100k' dataset contains 100,000 ratings from users on various movies and is used to evaluate the performance of recommendation algorithms in understanding user preferences and collaboration dynamics among content creators.

Dataset Card for ReDIAL Dataset Summary This is an adapted version of the original redial dataset, for supporting different tasks in our project RecBot. The redial dataset provides over 10,000 conversations centered around movie recommendations. It was released in the paper "Towards Deep Conversational Recommendations" at NeurIPS 2018. Supported Tasks and Leaderboards Sentiment Analysis: Use the SA config for sentiment analysis. Recommendation: Use the… See the full description on the dataset page: https://huggingface.co/datasets/recwizard/redial.

Epinions is a dataset that contains user-generated ratings and trust relationships, used to evaluate trust-based recommendation methods.

The 'Diginetica' dataset is a benchmark that contains user behavior sequences for session-based recommendation tasks, used to evaluate methods for predicting future user interests based on these anonymous interactions.

The Douban dataset contains user-item interactions and social relationships, and it is used to evaluate the effectiveness of recommendation models that incorporate high-order social relations.

Gowalla Dataset The Gowalla dataset, sourced from the Stanford Network Analysis Project (SNAP), contains user check-ins and social network information from the now-defunct location-based social networking platform Gowalla. Key features: Check-in data: records of user check-ins at various locations with timestamps and geographical coordinates (latitude, longitude). Social graph: user relationships represented as a graph, where edges denote friendships between users.… See the full description on the dataset page: https://huggingface.co/datasets/habedi/gowalla-dataset.

The 'Tmall' dataset is a large used car transaction dataset that contains resource-limitation characteristics and is used to evaluate user preferences in resource-limited recommendation tasks.

The 'Book-Crossing' dataset contains user ratings for books and is used to evaluate recommendation systems, particularly in the context of sequential recommendation tasks.

\

The 'Steam' dataset is a benchmark used to evaluate methods for product bundling in e-commerce by assessing the effectiveness of recommendations based on complementary item combinations.

The 'Beauty' dataset is a benchmark used to evaluate recommendation systems by providing data on user preferences and item attributes related to beauty products.

Foursquare is a dataset that contains user check-in data at various points of interest (POIs) and is used to evaluate next POI recommendation models.

Kuaishou is a large-scale short-video recommendation system dataset used to evaluate the effectiveness of generative retrieval methods in producing high-quality candidate recommendations under strict latency constraints.

The 'Twitter dataset' contains user feedback data used to evaluate the performance of recommendation algorithms, specifically in the context of recommending items to new users.

The Amazon products dataset contains product images and is used to evaluate the effectiveness of a deep-learning model for generating explainable recommendations based on image similarity.

BEIR is a benchmark dataset used to evaluate the performance of information retrieval systems across various tasks by providing a diverse set of query-document pairs.

Del.icio.us is a social bookmarking platform that contains user-tag-object tripartite graphs and is used to evaluate recommendation algorithms, particularly in addressing the cold-start problem in recommender systems.

The MS MARCO dataset is a benchmark that contains a large collection of query-document pairs used to evaluate document ranking models.

Dataset Card for "TripAdvisor" More Information needed

The 'Amazon Beauty' dataset is a benchmark that contains user interactions and product information specifically related to beauty products, used to evaluate the performance of sequential recommendation systems.

Goodreads is a dataset used to evaluate recommender systems, containing user ratings and reviews of books, which helps assess performance in diverse recommendation scenarios.

RecoGym is a simulation environment that contains a framework for modeling long-term rewards in recommender systems, used to evaluate the effectiveness of recommendations based on long-term business metrics rather than immediate rewards.

The 'Sports' dataset/benchmark contains user behavior data related to sports interests and is used to evaluate the effectiveness of recommendation systems in understanding and predicting user preferences through semantic reasoning.

The TRec dataset is an industrial benchmark used to evaluate large-scale recommendation systems, containing data that enables the assessment of ranking models' performance in real-world scenarios.

Amazon Review 2023 is an updated version of the Amazon Review 2018 dataset. This dataset mainly includes reviews (ratings, text) and item metadata (desc- riptions, category information, price, brand, and images). Compared to the pre- vious versions, the 2023 version features larger size, newer reviews (up to Sep 2023), richer and cleaner meta data, and finer-grained timestamps (from day to milli-second).

The Avazu dataset is used to evaluate click-through rate (CTR) prediction methods in computational advertising and recommender systems.

BibSonomy is a dataset that contains user-contributed bookmarks and associated tags, used to evaluate and improve tag recommendation algorithms by modeling cognitive processes in social tagging.

Criteo is a benchmark dataset used for evaluating click-through rate (CTR) prediction methods in computational advertising and recommender systems.

The DBLP dataset is a benchmark that contains bibliographic information on computer science publications and is used to evaluate recommendation algorithms.

The ML-100K dataset is a benchmark containing 100,000 ratings from users on movies, used to evaluate the performance of recommender systems.

Movielens-10M is a dataset that contains 10 million movie ratings from users, and it is used to evaluate the performance of recommender systems and methods for addressing confounding bias.

The 'MovieLens 25M' dataset contains 25 million ratings and tag applications from users on movies, and it is used to evaluate recommender systems by providing a large-scale benchmark for personalized movie recommendations.

The 'Pinterest' dataset is a benchmark that contains user-item interactions along with associated attributes, used to evaluate recommender systems, particularly in addressing cold-start problems and enhancing user-item interaction modeling.

The Polyvore dataset is a collection of fashion items used to evaluate fashion recommendation systems by assessing their ability to generate visually-collocated outfits.

The 'RetailRocket' dataset is a benchmark that contains user interaction sequences for session-based recommendation tasks, used to evaluate the effectiveness of recommendation models in predicting the next item a user will be interested in.

The 'RYM' dataset is a real-world benchmark used to evaluate recommendation systems, specifically focusing on aspects of recommendation accuracy and diversity.

The 'Yelp-2018' dataset is a benchmark used to evaluate recommender systems, containing user reviews and interactions that facilitate the assessment of recommendation quality and diversity.

The 'Yoochoose' dataset is a benchmark that contains anonymous user behavior sequences for session-based recommendation tasks, used to evaluate methods predicting users' future interests based on these sequences.

The 'YouTube' dataset/benchmark contains user interaction data with video recommendations and is used to evaluate the effectiveness of reinforcement learning methods in optimizing long-term user engagement in slate-based recommender systems.

Amazon Music is a dataset used to evaluate and improve music recommendation systems by comparing methods for diversifying recommendations, specifically through submodular and Jaccard approaches.

BERT-4Rec is a benchmark dataset used to evaluate sequential recommendation models by predicting a user's next interaction based on their historical behaviors.

The 'Ciao' dataset is a benchmark that contains user-generated reviews and ratings for various products, used to evaluate the performance of recommendation systems.

The 'Click-Through Rate (CTR) prediction benchmarks' consist of datasets used to evaluate the performance of models in predicting the likelihood that a user will click on an advertisement or recommendation.

The 'FilmTrust' dataset contains user ratings and trust relationships for movies and is used to evaluate recommendation algorithms, particularly in the context of addressing challenges posed by imperfect user behaviors.

The 'Games' dataset is a benchmark used to evaluate sequential recommendation systems, containing user interaction data specifically related to games.

The 'iFashion' dataset is a benchmark that contains a collection of fashion images used to evaluate the effectiveness of generative models in personalized outfit recommendation and composition.

The 'INSPIRED' dataset/benchmark contains conversational recommendation scenarios used to evaluate the performance of conversational recommender systems in delivering user-aligned recommendations.

The 'KuaiRec' dataset/benchmark contains user interaction data specifically designed to evaluate the performance of generative recommender systems in modeling long-term user behavioral patterns.

The 'Kwai' dataset is a benchmark used to evaluate multi-channel item-to-item retrieval systems in recommender systems, focusing on the effectiveness of trigger selection for enhancing retrieval performance.