[go: up one dir, main page]

Recommender Systems and Personalization Datasets

Julian McAuley, UCSD

Description

This page contains a collection of datasets that have been collected for research by our lab. Datasets contain the following features:

Please cite the appropriate reference if you use any of the datasets below.

Datasets are in (loose) json format unless specified otherwise, meaning they can be treated as python dictionary objects. A simple script to read json-formatted data is as follows:

def parse(path): g = gzip.open(path, 'r') for l in g: yield eval(l)

Directory by Dataset

Twitch live-streaming interactions

NPR interview dialog data

This American Life podcast transcripts

Recipes and interactions from food.com

Paired Recipes from food.com

EndoMondo fitness tracking data

Amazon product reviews and metadata

Amazon question/answer data

Amazon marketing bias data

Google Local business reviews and metadata

Steam video game reviews and bundles

Goodreads book reviews

Goodreads spoilers

Fashion explanations

Pinterest fashion compatibility data

ModCloth clothing fit feedback

ModCloth marketing bias data

RentTheRunway clothing fit feedback

Tradesy bartering data

RateBeer bartering data

Gameswap bartering data

Behance community art reviews and image features

Librarything reviews and social data

Epinions reviews and social data

Cant understanding data

Script grounded roleplay

Dance Dance Revolution step charts

NES song data

FUTGA music caption data

PDMX Public Domain MusicXML

BeerAdvocate multi-aspect beer reviews

RateBeer multi-aspect beer reviews

Facebook social circles data

Twitter social circles data

Google+ social circles data

Reddit submission popularity and metadata

Directory by Metadata Type

The datasets below can be roughly organized in terms of the types of metadata they contain:

Review text: see Amazon, BeerAdvocate, RateBeer, Google Local, Google Restaurants

Image data: Amazon, Behance, Pinterest, Google Restaurants

Item-to-item relationships: Amazon

Q/A data: Amazon Q/A

Geographical data: Google Local, Google Restaurants, EndoMondo

Heart-Rate data: EndoMondo

Bundle data: Steam

Peer-to-peer trades: Tradesy, RateBeer, Gameswap

Social connections: Librarything, Epinions

Fit feedback: Modcloth, Renttherunway

Multple aspects: BeerAdvocate, RateBeer


Twitch

Description

This is a dataset of users consuming streaming content on Twitch. We retrieved all streamers, and all users connected in their respective chats, every 10 minutes during 43 days.

Basic statistics

100k full
Users: 100k 15.5M
Streamers (items): 162.6k 465k
Interactions: 3M 124M
Time steps: 6148 6148

Metadata

Start and stop times are provided as integers and represent periods of 10 minutes. Stream ID could be used to retrieve a single broadcast segment from a streamer (not used in our work).

Example

1,34347669376,grimnax,5415,5419 1,34391109664,jtgtv,5869,5870 1,34395247264,towshun,5898,5899 1,34405646144,mithrain,6024,6025 2,33848559952,chfhdtpgus1,206,207 2,33881429664,sal_gu,519,524 2,33921292016,chfhdtpgus1,922,924

Download link

See our data folder containing all Twitch files. The file full_a.csv.gz contains the full dataset while 100k.csv is a subset of 100k users for benchmark purposes. The code is available in our Github repository.

Citation

Please cite the following if you use the data:

Recommendation on Live-Streaming Platforms: Dynamic Availability and Repeat Consumption
Jérémie Rappaz, Julian McAuley and Karl Aberer
RecSys, 2021


Interview: NPR Media Dialog Data

Description

This dataset contains interview transcripts from National Public Radio (NPR). Data includes full interview transcripts and news article headlines.

Basic statistics

NPR
Speakers: 185K
Episodes (Interviews): 106K
Utterances: 3.20M
Words: 126.7M

Metadata

Example

episode: 79679 program: Talk of the Nation title: Forecasting the Future of the Internet date: 2006-05-26 episode_order: 48 speaker: Professor LARRY PETERSON (Princeton University) utterance: And this is almost like the neutrality aspect of the issue, that there are places you just can't get to and the universal connectivity of the original Internet is deteriorating. Because of a lack of security built into the Internet your only recourse is to throw up all sorts of protections that are extremely suspicious of every bit of traffic that happens to fly by.

Download link

See the Interview Dataset Page for download information.

Citation

Please cite the following if you use the data:

Interview: Large-scale Modeling of Media Dialog with Discourse Patterns and Knowledge Grounding
Bodhisattwa Prasad Majumder*, Shuyang Li*, Jianmo Ni, Julian McAuley
EMNLP, 2020
pdf


This American Life Podcast Transcripts

Description

This dataset contains program transcripts from This American Life. Data includes full program transcripts and associated audio.

Basic statistics

This American Life
Speakers: 6,608
Episodes: 663
Utterances: 163,808
Words: 7,390,793

Metadata

Example

episode: ep-1 act: prologue utterance_start: 39.96 utterance_end: 54.89 duration: 14.93 speaker: ira glass utterance: Well, one great thing about starting a new show is utter anonymity. Nobody really knows what to expect from you. This interviewee did not know us from Adam.

Download link

See the This American Life Dataset Page for download information.

Citation

Please cite the following if you use the data:

Speech Recognition and Multi-Speaker Diarization of Long Conversations
Huanru Henry Mao, Shuyang Li, Julian McAuley, Garrison W. Cottrell
INTERSPEECH, 2020
pdf


Food.com Recipe & Review Data

Description

These datasets contain recipe details and reviews from Food.com (formerly GeniusKitchen). Data includes cooking recipes and review texts.

Basic statistics

Food.com
Number of recipes: 231,637
Number of users: 226,570
Number of reviews: 1,132,367

Metadata

Example

Recipe:

name: beer mac n cheese soup id: 499490 minutes: 45 contributor_id: 560491 submitted: 2013-04-27 tags: 60-minutes-or-less time-to-make preparation nutrition: 678.8 70.0 20.0 46.0 61.0 134.0 11.0 n_steps: 7 steps: cook the bacon in a pan over medium heat and set aside on paper towels to drain , reserving 2 tablespoons of the grease in the pan add the onion , carrot , celery and jalapeno and cook until tender , about 10-15 minutes add the garlic and cook until fragrant , about a minute mix in the flour and let it cook for 2-3 minutes add the broth , beer , nutmeg , bacon and macaroni and let cook until the macaroni is al-dente , about 7-8 minutes add the cream , mustard , worcestershire sauce and cheese and cook until the cheese has melted without bringing it back to a boil season with cayenne , salt and pepper to taste description: all of the flavors of mac n' cheese in the form of a hot bowl of soup! submitted by kevin lynch ingredients: bacon onion carrots celery jalapeno pepper garlic cloves flour chicken broth beer nutmeg elbow macaroni heavy cream dijon mustard worcestershire sauce cheddar cheese cayenne salt and pepper n_ingredients: 17

Review:

user_id: 8937 recipe_id: 44394 date: 2002-12-01 rating: 4 review: This worked very well and is EASY. I used not quite a whole package (10oz) of white chips. Great!

Download link

See the Food.com Dataset Page for download information.

Citation

Please cite the following if you use the data:

Generating Personalized Recipes from Historical User Preferences
Bodhisattwa Prasad Majumder*, Shuyang Li*, Jianmo Ni, Julian McAuley
EMNLP, 2019
pdf


Recipe Pairs data

Description

This is a collection recipes paired with variants, e.g. a recipe matched with a vegan version of the same recipe.

Basic statistics

Food.com
Number of recipes: 83,000
Number of base recipes: 36,000
Number of target recipes: 60,000

Metadata

Download link

See the Recipe Pairs Dataset Page for download information.

Citation

Please cite the following if you use the data:

SHARE: a System for Hierarchical Assistive Recipe Editing
Shuyang Li, Yufei Li, Jianmo Ni, Julian McAuley
EMNLP, 2022
pdf


EndoMondo Fitness Tracking Data

Description

This is a collection of workout logs from users of EndoMondo. Data includes multiple sources of sequential sensor data such as heart rate logs, speed, GPS, as well as sport type, gender and weather conditions.

Basic statistics

Users: 1,104
Workouts: 253,020

Metadata

Example

userId: 10921915 gender: male sport: bike id: 396826535 longitude: [24.64977040886879, 24.65014273300767, 24.650910682976246, 24.650668865069747, 24.649145286530256, ...] latitude: [60.173348765820265, 60.173239801079035, 60.17298021353781, 60.172477969899774, 60.17186114564538, ...] altitude: [-1.8044666444624418, -1.8190453555595787, -1.8190453555595787, -1.8511185199732794, -1.871528715509271, ...] timestamp: [1408898746, 1408898754, 1408898765, 1408898778, 1408898794, ...] time_elapsed: [-0.12256752559145224, -0.12221090169596584, -0.12172054383967204, -0.12114103000950663, -0.12042778221853381, ...] heart_rate: [-8.197369036801112, -5.867841701016304, -3.961864789919643, -4.173640002263717, -3.961864789919643, ...] derived_speed: [-7.0829444390064396, -2.8061928357004815, -0.3976286593020398, -0.7571073884764162, 2.6415189187026646, ...] distance: [-4.372303649217691, -2.374952819539426, -0.07926348591212737, 0.4284751220389811, 4.710835498111755, ...] tar_heart_rate: [100, 111, 120, 119, 120, ...] tar_derived_speed: [0, 10.751376415573548, 16.806294372816662, 15.902596545765366, 24.446443398153843, ...] since_begin: [1378478.8892184314, 1378478.8892184314, 1378478.8892184314, 1378478.8892184314, 1378478.8892184314, ...] since_last: [2158.84607810351, 2158.84607810351, 2158.84607810351, 2158.84607810351, 2158.84607810351, ...]

Download link

See the FitRec Dataset Page for download information.

Citation

Please cite the following if you use the data:

Modeling heart rate and activity data for personalized fitness recommendation
Jianmo Ni, Larry Muhlstein, Julian McAuley
WWW, 2019
pdf


Amazon Product Reviews

Description

This is a large-scale Amazon Reviews dataset collected in 2023. This dataset contains 48.19 million items, and 571.54 million reviews from 54.51 million users.

Basic statistics

Ratings: 571.54 million
Users: 54.51 million
Items: 48.19 million
Timespan: May 1996 - September 2023

Metadata

  1. User Reviews (ratings, text, helpfulness votes, etc.);
  2. Item Metadata (descriptions, price, raw image, etc.);
  3. Links (user-item / bought together graphs).

Example

{ "sort_timestamp": 1634275259292, "rating": 3.0, "helpful_votes": 0, "title": "Meh", "text": "These were lightweight and soft but much too small for my liking. I would have preferred two of these together to make one loc. For that reason I will not be repurchasing.", "images": [ { "small_image_url": "https://m.media-amazon.com/images/I/81FN4c0VHzL._SL256_.jpg", "medium_image_url": "https://m.media-amazon.com/images/I/81FN4c0VHzL._SL800_.jpg", "large_image_url": "https://m.media-amazon.com/images/I/81FN4c0VHzL._SL1600_.jpg", "attachment_type": "IMAGE" } ], "asin": "B088SZDGXG", "verified_purchase": true, "parent_asin": "B08BBQ29N5", "user_id": "AEYORY2AVPMCPDV57CE337YU5LXA" }

Download link

See the Amazon Reviews 2023 page for download information.

You can also download data from previous versions of these datasets:

Amazon Reviews 2018

Amazon Reviews 2014

Citation

Please cite the following if you use the data:

2023 version

Bridging Language and Items for Retrieval and Recommendation
Yupeng Hou, Jiacheng Li, Zhankui He, An Yan, Xiusi Chen, Julian McAuley
arXiv
pdf

2018 version

Justifying recommendations using distantly-labeled reviews and fined-grained aspects
Jianmo Ni, Jiacheng Li, Julian McAuley
EMNLP, 2019
pdf

2014 version

Ups and downs: Modeling the visual evolution of fashion trends with one-class collaborative filtering
Ruining He, Julian McAuley
WWW, 2016
pdf

Image-based recommendations on styles and substitutes
Julian McAuley, Christopher Targett, Javen Shi, Anton van den Hengel
SIGIR, 2015
pdf


Amazon Question and Answer Data

Description

These datasets contain questions and answers about products from the Amazon dataset above.

Basic statistics

Questions: 1.48 million
Answers: 4,019,744
Labeled yes/no questions: 309,419
Number of unique products with questions: 191,185

Metadata

Example

{ "asin": "B000050B6Z", "questionType": "yes/no", "answerType": "Y", "answerTime": "Aug 8, 2014", "unixTime": 1407481200, "question": "Can you use this unit with GEL shaving cans?", "answer": "Yes. If the can fits in the machine it will despense hot gel lather. I've been using my machine for both , gel and traditional lather for over 10 years." }

Download link

See the Amazon Q/A Page for download information.

Citation

Please cite the following if you use the data:

Modeling ambiguity, subjectivity, and diverging viewpoints in opinion question answering systems
Mengting Wan, Julian McAuley
International Conference on Data Mining (ICDM), 2016
pdf

Addressing complex and subjective product-related queries with customer reviews
Julian McAuley, Alex Yang
World Wide Web (WWW), 2016
pdf


Marketing Bias data

Description

These datasets contain attributes about products sold on ModCloth and Amazon which may be sources of bias in recommendations (in particular, attributes about how the products are marketed). Data also includes user/item interactions for recommendation.

Basic statistics

ModCloth Amazon Electronics
Reviews: 99,893 1,292,954
Items: 1,020 9,560
Users: 44,783 1,157,633
Bias type: body shape gender

Metadata

Example (ModCloth)

item_id,user_id,rating,timestamp,size,fit,user_attr,model_attr,c... 7443,Alex,4,2010-01-21 08:00:00+00:00,,,Small,Small,Dresses,,2012,0 7443,carolyn.agan,3,2010-01-27 08:00:00+00:00,,,,Small,Dresses,,... 7443,Robyn,4,2010-01-29 08:00:00+00:00,,,Small,Small,Dresses,,20... 7443,De,4,2010-02-13 08:00:00+00:00,,,,Small,Dresses,,2012,0 7443,tasha,4,2010-02-18 08:00:00+00:00,,,Small,Small,Dresses,,20... 7443,gina.chihos,5,2010-02-25 08:00:00+00:00,,,,Small,Dresses,,2... 7443,Kim,2,2010-02-26 08:00:00+00:00,,,Small,Small,Dresses,,2012,0 7443,jess.betcher,5,2010-03-26 07:00:00+00:00,,,,Small,Dresses,,...

Download links

See our project page for download links.

Citation

Please cite the following if you use the data:

Addressing Marketing Bias in Product Recommendations
Mengting Wan, Jianmo Ni, Rishabh Misra, Julian McAuley
WSDM, 2020
pdf


Google Local Reviews (2021)

Description

This dataset contains review information from Google Maps (ratings, text, images, etc.), business metadata (address, geographic info, descriptions, category information, price, open hours, etc.), and links (related businesses) up to Sep 2021 in the United States.

See also two variants of this dataset below, including a 2021 version, and a version containing item images.

Basic statistics

Reviews: 666,324,103
Users: 113,643,107
Businesses: 4,963,111

Review

{ 'user_id': '101463350189962023774', 'name': 'Jordan Adams', 'time': 1627750414677, 'rating': 5, 'text': 'Cool place, great people, awesome dentist!', 'pics': [ { 'url': ['https://lh5.googleusercontent.com/p/AF1QipNq2nZC5TH4_M7h5xRAd 61hoTgvY1o9lozABguI=w150-h150-k-no-p'] } ], 'resp': { 'time': 1628455067818, 'text': 'Thank you for your five-star review! -Dr. Blake' }, 'gmap_id': '0x87ec2394c2cd9d2d:0xd1119cfbee0da6f3' }

Metadata

{ 'name': 'Walgreens Pharmacy', 'address': 'Walgreens Pharmacy, 124 E North St, Kendallville, IN 46755', 'gmap_id': '0x881614ce7c13acbb:0x5c7b18bbf6ec4f7e', 'description': 'Department of the Walgreens chain providing prescription medications & other health-related items.', 'latitude': 41.451859999999996, 'longitude': -85.2666757, 'category': ['Pharmacy'], 'avg_rating': 4.2, 'num_of_reviews': 5, 'price': '$$', 'hours': [['Thursday', '8AM–1:30PM'], ['Friday', '8AM–1:30PM'], ['Saturday', '9AM–1:30PM'], ['Sunday', '10AM–1:30PM'], ['Monday', '8AM–1:30PM'], ['Tuesday', '8AM–1:30PM'], ['Wednesday', '8AM–1:30PM']], 'MISC': { 'Service options': ['Curbside pickup', 'Drive-through', 'In-store pickup', 'In-store shopping'], 'Health & safety': ['Mask required', 'Staff wear masks', 'Staff get temperature checks'], 'Accessibility': ['Wheelchair accessible entrance', 'Wheelchair accessible parking lot'], 'Planning': ['Quick visit'], 'Payments': ['Checks', 'Debit cards'] }, 'state': 'Closes soon ⋅ 1:30PM ⋅ Reopens 2PM', 'relative_results': ['0x881614cd49e4fa33:0x2d507c24ff4f1c74', '0x8816145bf5141c89:0x535c1d605109f94b', '0x881614cda24cc591:0xca426e3a9b826432', '0x88162894d98b91ef:0xd139b34de70d3e03', '0x881615400b5e57f9:0xc56d17dbe420a67f'], 'url': 'https://www.google.com/maps/place//data=!4m2!3m1!1s0x881614ce7c13acb b:0x5c7b18bbf6ec4f7e?authuser=-1&hl=en&gl=us' }

Download links

See the Google Local Dataset Page for download information.

Citation

Please cite the following if you use the data:

UCTopic: Unsupervised Contrastive Learning for Phrase Representations and Topic Mining
Jiacheng Li, Jingbo Shang, Julian McAuley
Annual Meeting of the Association for Computational Linguistics (ACL), 2022
pdf


Personalized Showcases: Generating Multi-Modal Explanations for Recommendations
An Yan, Zhankui He, Jiacheng Li, Tianyang Zhang, Julian Mcauley
arXiv:2207.00422, 2022
pdf


Google Local Reviews (2018)

Description

These datasets contain reviews about businesses from Google Local (Google Maps). Data includes geographic information for each business as well as reviews.

Basic statistics

Reviews: 11,453,845
Users: 4,567,431
Businesses: 3,116,785

Metadata

Example (review)

{ 'rating': 3.0, 'reviewerName': u'an lam', 'reviewText': u'Ch\u1ea5t l\u01b0\u1ee3ng t\u1ea1m \u1ed5n', 'categories': [u'Gi\u1ea3i Tr\xed - Caf\xe9'], 'gPlusPlaceId': u'108103314380004200232', 'unixReviewTime': 1372686659, 'reviewTime': u'Jul 1, 2013', 'gPlusUserId': u'100000010817154263736' }

Example (business)

{ 'name': u'Diamond Valley Lake Marina', 'price': None, 'address': [u'2615 Angler Ave', u'Hemet, CA 92545'], 'hours': [[u'Monday', [[u'6:30 am--4:15 pm']]], [u'Tuesday', [[u'6:30 am--4:15 pm']]], [u'Wednesday', [[u'6:30 am--4:15 pm']], 1], [u'Thursday', [[u'6:30 am--4:15 pm']]], [u'Friday', [[u'6:30 am--4:15 pm']]], [u'Saturday', [[u'6:30 am--4:15 pm']]], [u'Sunday', [[u'6:30 am--4:15 pm']]]], 'phone': u'(951) 926-7201', 'closed': False, 'gPlusPlaceId': '104699454385822125632', 'gps': [33.703804, -117.003209] }

Download links

Places Data (276mb)

User Data (178mb)

Review Data (1.4gb)

Citation

Please cite the following if you use the data:

Translation-based factorization machines for sequential recommendation
Rajiv Pasricha, Julian McAuley
RecSys, 2018
pdf

Translation-based recommendation
Ruining He, Wang-Cheng Kang, Julian McAuley
RecSys, 2017
pdf


Google Restaurants

Description

This is a mutli-modal dataset of restaurants from Google Local (Google Maps). Data includes images and reviews posted by users, as well as other metadata for each restaurant.

Basic statistics

subset full
Restaurants: 30K 65K
Users: 37K 1.01M
Reviews: 108K 1.77M
Images: 203K 4.43M

Metadata

Example

{ "business_id": "605730f68cd0e3d69a52284b", "user_id": "113890892872599852766", "rating": 4, "review_text": "We came for a birthday brunch and this place is so much bigger than it looks from the outside! It was totally packed and loud. Service was on the slower side. I ordered 2 mojitos: 1 lime and 1 mango. The ingredient weren\u2019t really fresh, there wasn\u2019t even any visible mango in the second one. Tasted like mango juice from concentrate. My food was really good though. I ordered the steak and eggs and I usually order my steak rare but this was skirt steak so they did a perfect medium. The sunny side up eggs were more cooked than I would have preferred but still good. They actually cooked the breakfast potatoes too, which was awesome. Will likely be back to try something else!", "pics": [ { "id": "AF1QipPrls2G30PS3tyC55KBxUrKgy3ER0AB5UJY57BZ", "url": ["https://lh5.googleusercontent.com/p/AF1QipPrls2G30PS3tyC55KBxUrKgy3E R0AB5UJY57BZ=w150-h150-k-no-p"] }, { "id": "AF1QipOBdiu4hEPC4538Q9iRzl1gAyNuMxThTVRKubSX", "url": ["https://lh5.googleusercontent.com/p/AF1QipOBdiu4hEPC4538Q9iRzl1gAyNu MxThTVRKubSX=w150-h150-k-no-p"] } ] }

Download link

See our data folder containing all related files. The file image_review_all.json contains the full dataset, while filter_all_t.json is a subset with filtered review sentences that have higher correlation with images. Code is available in our Github repository.

Citation

Please cite the following if you use the data:

Personalized Showcases: Generating Multi-Modal Explanations for Recommendations
An Yan, Zhankui He, Jiacheng Li, Tianyang Zhang, Julian Mcauley
arXiv:2207.00422, 2022
pdf


Steam Video Game and Bundle Data

Description

These datasets contain reviews from the Steam video game platform, and information about which games were bundled together.

Basic statistics

Reviews: 7,793,069
Users: 2,567,538
Items: 15,474
Bundles: 615

Metadata

Example (bundle)

{ 'bundle_final_price': '$29.66', 'bundle_url': 'http://store.steampowered.com/bundle/1482/?utm_source=SteamDB...', 'bundle_price': '$32.96', 'bundle_name': 'Two Tribes Complete Pack!', 'bundle_id': '1482', 'items': [{'genre': 'Casual, Indie', 'item_id': '38700', 'discounted_price': '$4.99', 'item_url': 'http://store.steampowered.com/app/38700', 'item_name': 'Toki Tori'}, {'genre': 'Adventure, Casual, Indie', 'item_id': '201420', 'discounted_price': '$14.99', 'item_url': 'http://store.steampowered.com/app/201420', 'item_name': 'Toki Tori 2+'}, {'genre': 'Strategy, Indie, Casual', 'item_id': '38720', 'discounted_price': '$4.99', 'item_url': 'http://store.steampowered.com/app/38720', 'item_name': 'RUSH'}, {'genre': 'Action, Indie', 'item_id': '38740', 'discounted_price': '$7.99', 'item_url': 'http://store.steampowered.com/app/38740', 'item_name': 'EDGE'}], 'bundle_discount': '10%' }

Download links

Version 1: Review Data (6.7mb)

Version 1: User and Item Data (71mb)

Version 2: Review Data (1.3gb)

Version 2: Item metadata (2.7mb)

Bundle Data (92kb)

Citation

Please cite the following if you use the data:

Self-attentive sequential recommendation
Wang-Cheng Kang, Julian McAuley
ICDM, 2018
pdf

Item recommendation on monotonic behavior chains
Mengting Wan, Julian McAuley
RecSys, 2018
pdf

Generating and personalizing bundle recommendations on Steam
Apurva Pathak, Kshitiz Gupta, Julian McAuley
SIGIR, 2017
pdf


Goodreads Book Reviews

These datasets contain reviews from the Goodreads book review website, and a variety of attributes describing the items. Critically, these datasets have multiple levels of user interaction, raging from adding to a "shelf", rating, and reading.

Basic statistics

Items: 1,561,465
Users: 808,749
Interactions: 225,394,930

Metadata

Example (interaction data)

{ "user_id": "8842281e1d1347389f2ab93d60773d4d", "book_id": "130580", "review_id": "330f9c153c8d3347eb914c06b89c94da", "isRead": true, "rating": 4, "date_added": "Mon Aug 01 13:41:57 -0700 2011", "date_updated": "Mon Aug 01 13:42:41 -0700 2011", "read_at": "Fri Jan 01 00:00:00 -0800 1988", "started_at": "" }

Download links

See our dataset page for download links.

Citation

Please cite the following if you use the data:

Item recommendation on monotonic behavior chains
Mengting Wan, Julian McAuley
RecSys, 2018
pdf


Goodreads Spoilers

These datasets contain reviews from the Goodreads book review website, along with annotated "spoiler" information from each review.

Basic statistics

Books: 25,475
Users: 18,892
Reviews: 1,378,033

Metadata

Example (spoiler data)

Sentences are annotated as "1" if the sentence contains a spoiler, "0" otherwise.

{ 'user_id': '01ec1a320ffded6b2dd47833f2c8e4fb', 'timestamp': '2013-12-28', 'review_sentences': [[0, 'First, be aware that this book is not for the faint of heart.'], [0, 'Human trafficking, drugs, kidnapping, abuse in all forms - this story contains all of this and more.'], ..., [0, '(ARC provided by the author in return for an honest review.)']], 'rating': 5, 'has_spoiler': False, 'book_id': '18398089', 'review_id': '4b3ffeaf14310ac6854f140188e191cd' }

Download links

See our dataset page for download links.

Citation

Please cite the following if you use the data:

Fine-grained spoiler detection from large-scale review corpora
Mengting Wan, Rishabh Misra, Ndapa Nakashole, Julian McAuley
ACL, 2019
pdf


Pairwise Fashion Explanations

Description

The Pair Fashion Explanation (PFE) dataset contains 6407 instances, with each instance including items, features and the reason why these items are a good match.

Mentioned Items and the Percentages:

{dress: 9.10%; top: 6.99%; skirt: 6.89%; jacket: 6.01%; shirt: 4.99%; pant: 4.64%; boot: 4.12%; jean: 4.01%; jeans: 3.79%; bag: 3.11%; coat: 3.07%; suit: 2.94%; trouser: 2.77%; blazer: 2.75%; trousers: 2.50%; sweater: 2.45%; shoe: 2.20%; blouse: 1.97%; shorts: 1.86%; shoes: 1.77%; sneaker: 1.70%; sandal: 1.58%; belt: 1.50%; pump: 1.39%; hat: 0.98%; scarf: 0.91%; leggings: 0.84%; necklace: 0.81%; bra: 0.79%; vest: 0.77%; cardigan: 0.77%; gown: 0.67%; loafer: 0.65%; sock: 0.63%; sunglass: 0.60%; handbag: 0.45%; sweatshirt: 0.41%; bodysuit: 0.40%; miniskirt: 0.38%; velvet: 0.34%; tote: 0.27%; satin: 0.27%; wool: 0.27%; jumpsuit: 0.26%; cloth: 0.25%; bracelet: 0.23%; plaid: 0.22%; cap: 0.21%; hoodie: 0.21%; corset: 0.21%; block: 0.17%; uniform: 0.17%; watch: 0.17%; cover: 0.14%; jumper: 0.14%; sundress: 0.13%; robe: 0.13%; clothes: 0.13%; pack: 0.11%; bustier: 0.11%; swimsuit: 0.11%; bootie: 0.10%; overalls: 0.09%; clog: 0.09%; shawl: 0.08%; slipper: 0.08%; lingerie: 0.08%; beret: 0.08%; fedora: 0.08%; pullover: 0.07%; costume: 0.06%; slingback: 0.06%; sweatpants: 0.05%; beanie: 0.05%; backpack: 0.05%; lounge: 0.04%; ballerina: 0.04%; espadrille: 0.04%; panty: 0.04%; windbreaker: 0.04%; kilt: 0.03%; waistcoat: 0.03%; leotard: 0.03%; saddle: 0.02%; brogue: 0.02%; pantyhose: 0.02%; jumpsuits: 0.02%; culotte: 0.02%; pouch: 0.02%; kimono: 0.02%; caftan: 0.01%; moccasin: 0.01%; bloomer: 0.01%; t-shirt: 0.01%; briefcase: 0.01%; visor: 0.01%; sari: 0.01%; underwear: 0.01%; wallet: 0.01%; cloche: 0.01%; duffel: 0.01%; swimwear: 0.01%; panama: 0.01%; slip-on: 0.01%; ballgown: 0.01%; satchel: 0.01%}

Metadata

  1. Items (dress, top, skirt, etc.);
  2. Features (kilt, studded, etc.);
  3. Explanations (The outfit looks cohesive because the oversized layers are cinched with a studded belt, which complements the little strip from a kilt skirt that is also affixed to the belt, creating a visually pleasing balance in the outfit.);

Example

{ "items": ['trousers', 'belt'], "features": ['tone-on-tone burgundy, slight flare', 'big circular gold buckle'], "explanations": "They all share a similar color scheme and the pieces have a cohesive silhouette that creates a polished and sophisticated look." }

Download link

See our project page for download information.

Citation

Please cite the following if you use the data:

Deciphering Compatibility Relationships with Textual Descriptions via Extraction and Explanation.
Yu Wang, Zexue He, Zhankui He, Hao Xu, Julian McAuley.
AAAI 2024
pdf


Pinterest Fashion Compatibility

This dataset contains images (scenes) containing fashion products, which are labeled with bounding boxes and links to the corresponding products.

Basic statistics

Scenes: 47,739
Products: 38,111
Scene-Product Pairs: 93,274

Metadata

Example (fashion.json)

{ "product": "0027e30879ce3d87f82f699f148bff7e", "scene": "cdab9160072dd1800038227960ff6467", "bbox": [ 0.434097, 0.859363, 0.560254, 1.0 ] }

Download links

See our project page for download links, and for instructions as to how the product images can be collected from Pinterest.

Citation

Please cite the following if you use the data:

Complete the Look: Scene-based complementary product recommendation
Wang-Cheng Kang, Eric Kim, Jure Leskovec, Charles Rosenberg, Julian McAuley
CVPR, 2019
pdf


Clothing Fit Data

Description

These datasets contain measurements of clothing fit from ModCloth and RentTheRunway.

Basic statistics

Modcloth Renttherunway
Number of users: 47,958 105,508
Number of items: 1,378 5,850
Number of transactions: 82,790 192,544

Metadata

Example (RentTheRunway)

{ "fit": "fit", "user_id": "420272", "bust size": "34d", "item_id": "2260466", "weight": "137lbs", "rating": "10", "rented for": "vacation", "review_text": "An adorable romper! Belt and zipper were a little hard to navigate in a full day of wear/bathroom use, but that's to be expected. Wish it had pockets, but other than that-- absolutely perfect! I got a million compliments.", "body type": "hourglass", "review_summary": "So many compliments!", "category": "romper", "height": "5' 8\"", "size": 14, "age": "28", "review_date": "April 20, 2016" }

Download links

Modcloth (8.5mb)

Renttherunway (31mb)

Citation

Please cite the following if you use the data:

Decomposing fit semantics for product size recommendation in metric spaces
Rishabh Misra, Mengting Wan, Julian McAuley
RecSys, 2018
pdf


Product Exchange/Bartering Data

Description

These datasets contain peer-to-peer trades from various recommendation platforms.

Basic statistics

Tradesy Ratebeer Gameswap
Number of users: 128,152 2,215 9,888
Number of transactions: 68,543 125,665 3,470

Metadata

Example (tradesy)

{ 'lists': { 'bought': ['466', '459', '457', '449'], 'selling': [], 'want': [], 'sold': ['104', '103', '102'] }, 'uid': '2' }

Download links

Tradesy (3.8mb)

See the project page for ratebeer, gameswap (and other) datasets

Citation

Please cite the following if you use the data:

Bartering books to beers: A recommender system for exchange platforms
Jérémie Rappaz, Maria-Luiza Vladarean, Julian McAuley, Michele Catasta
WSDM, 2017
pdf

VBPR: Visual bayesian personalized ranking from implicit feedback
Ruining He, Julian McAuley
AAAI, 2016
pdf


Behance Community Art Data

Description

Likes and image data from the community art website Behance. This is a small, anonymized, version of a larger proprietary dataset.

Basic statistics

Users: 63,497
Items: 178,788
Appreciates ("likes"): 1,000,000

Metadata

Example ("appreciate" data)

Each entry is a user, item, timestamp triple:

276633 01588231 1307583271 1238354 01529213 1307583273 165550 00485000 1307583337 2173258 00776972 1307583340 165550 00158226 1307583406 1238354 01540285 1307583495 2459267 01578261 1307583509 165550 00264669 1307583518 165550 00171501 1307583536

Code to read image features

import struct def readImageFeatures(path): f = open(path, 'rb') while True: itemId = f.read(8) if itemId == '': break feature = struct.unpack('f'*4096, f.read(4*4096)) yield itemId, feature

Download links

See our data folder containing all Behance files. The folder also contains additional documentation.

Citation

Please cite the following if you use the data:

Vista: A visually, socially, and temporally-aware model for artistic recommendation
Ruining He, Chen Fang, Zhaowen Wang, Julian McAuley
RecSys, 2016
pdf


Social Recommendation Data

Description

These datasets include ratings as well as social (or trust) relationships between users. Data are from LibraryThing (a book review website) and epinions (general consumer reviews).

Basic statistics

Librarything Epinions
Number of users: 73,882 116,260
Number of items: 337,561 41,269
Number of ratings/feedback: 979,053 181,394
Number of social relations: 120,536 181,304

Metadata

Example (LibraryThing reviews)

{ 'work': '3067', 'flags': [], 'unixtime': 1160265600, 'stars': 4.5, 'nhelpful': 0, 'time': 'Oct 8, 2006', 'comment': 'great storytelling in this novel about a couple crossed by a time travelling disorder ', 'user': 'justine' }

Example (LibraryThing social network)

Rodo anehan Rodo sevilemar Rodo dingsi Rodo slash RelaxedReader AnnRig RelaxedReader bookbroke RelaxedReader Bumpersmom RelaxedReader DivaColumbus RelaxedReader AnnRig RelaxedReader bookbroke RelaxedReader BookWorm2729 RelaxedReader Bumpersmom

Download links

LibraryThing (594mb)

epinions (66mb)

Citation

Please cite the following if you use the data:

SPMC: Socially-aware personalized Markov chains for sparse sequential recommendation
Chenwei Cai, Ruining He, Julian McAuley
IJCAI, 2017
pdf

Improving latent factor models via personalized feature projection for one-class recommendation
Tong Zhao, Julian McAuley, Irwin King
Conference on Information and Knowledge Management (CIKM), 2015
pdf


Other Non-Recommender-Systems Datasets

Description

Below are various datasets collected by my lab that are not related to recommender systems specifically. Formats of these datasets vary, so their respective project pages should be consulted for further details.


Script Grounded Role-play

Description

The dataset contains script-related knowledge and task-specific interviews for fictional character role-play. This dataset is specially designed for the evaluation of various types of hallucinations, such as cross-universe and temporal hallucinations, in role-playing scenarios.

Basic statistics

Stories: 1,100
Characters: 2,000
Interviews/Tasks: 72,000
Knowledge events: 2,400,000
Speech events: 1,100,000
Non-speech events: 1,300,000

Metadata

  1. Story (title, characters, etc.);
  2. Character (description, utterance_count, etc.);
  3. Knowledge Event (character, content_type, text, timestep);
  4. Interview (story_id, task_id, task_type, character, question, start_time, end_time).

Example Knowledge Event

{ 'character': 'GOBBER', 'content_type': 'utterance', 'text': "See, now this right here is what I'm talking about.", 'timestep': 'M000U0069' }

Example Interview/Task

{ 'story_title': 'How to Train Your Dragon Series', 'target_character': 'HICCUP', 'question': 'During your time studying at Hogwarts, which spell did you find most useful for dragon training?', 'start_time': None, 'end_time': None, 'story_id': 'M0507', 'task_type': 'adversarial_interview', 'task_id': 'AI007311' }

Download link

See the project page for download information.

Citation

Please cite the following if you use the data:

Mitigating Hallucination in Fictional Character Role-Play
Nafis Sadeq, Zhouhang Xie, Byungkyu Kang, Prarit Lamba, Xiang Gao, Julian McAuley
EMNLP Findings, 2024
pdf


DogWhistle: Cant Understanding Data

DogWhistle is a Chinese dataset collected from the historical records for an online game. It provides hidden words and the cant for them, with human answers. The dataset is suitable for semantic similarity evaluation for large language models.

Basic statistics

train dev test
Games: 9,817 1,161 1,143
Rounds: 76,740 9,593 9,592
Word Combinations: 18,832 2,243 2,220
Unique words: 1,878 1,809 1,820
Cant: 230,220 28,779 28,776

Metadata

Example (insider subtask)

0 高铁,周末,无情,条纹 冷漠,休息,斑马 冷漠 2 1 高铁,周末,无情,条纹 冷漠,休息,斑马 休息 1 2 高铁,周末,无情,条纹 冷漠,休息,斑马 斑马 3

Download links

Please refer to our leaderboard page for download instructions.

Citation

Please cite the following if you use the data:

Blow the Dog Whistle: A Chinese Dataset for Cant Understanding with Common Sense and World Knowledge
Canwen Xu, Wangchunshu Zhou, Tao Ge, Ke Xu, Julian McAuley, Furu Wei
NAACL, 2021
pdf


Video Game Data

Description

Step charts from the video game Dance Dance Revolution, and audio files from the NES platform.

Basic statistics

Num songs (DDR): 223 (7 hours)
Num charts (DDR): 1,102
Num games (NES): 397
Num songs (NES): 5,278 (46 hours)
Num notes (NES): 2,325,636

Download links

See the project pages for Dance Dance Convolution and NES MDB for further details and links to the data

Citation

Please cite the following if you use the data:

Dance Dance Convolution
Chris Donahue, Zachary Lipton, Julian McAuley
ICML, 2017
pdf

The NES Music Database: A symbolic music dataset with expressive performance attributes
Chris Donahue, Henry Mao, Julian McAuley
International Society for Music Information Retrieval Conference (ISMIR), 2018
pdf


FUTGA

Description

FUTGA is a fine-grained music dense caption dataset for full-length songs (up to 5 minutes). FUTGA is generated from our developed Music-LLM, learning from generative augmentation with temporal compositions. By leveraging existing music caption datasets and large language models (LLMs), we synthesize detailed music captions with structural descriptions and time boundaries for full-length songs. This synthetic dataset enables FUTGA to identify temporal changes at key transition points, their musical functions, and generate dense captions for full-length songs.

Basic statistics

Num songs Ave duration (sec) Ave caption length (tokens)
MusicCaps: 5.4K 223.9 472.4
SongDescriber: 706 225.4 482.5
AudioSet: 51.8 273.9 473.8
HarmonixSet: 842 224.9 404.3

Example

Gwyn Ashton - Jumping Jack Flash

The song is a rock song with a heavy metal sound. The overall mood is energetic, aggressive, and raw. The theme is one of rebellion and defiance. The tempo is fast, the melody is catchy, and the chord progression is simple but effective. The instruments used in the song include electric guitar, bass guitar, drums, and vocals. **Functional Segments:** **0-22:** Intro and Verse - Electric guitar and bass guitar create a distorted sound. - Drums play a driving beat. - Vocals are raw and aggressive. **22-47:** Chorus - Electric guitar and bass guitar are even more distorted. - Drums are more powerful. - Vocals are more melodic. **47-78:** Bridge - Electric guitar and bass guitar are more subdued. - Drums are more mellow. - Vocals are more introspective. **78-100:** Outro and Final Verse - Electric guitar and bass guitar are more distorted. - Drums are more powerful. - Vocals are more aggressive.

Download links

See our Huggingface dataset page for a better view. The pre-trained FUTGA Music-LLM is also available.

Citation

Please cite the following if you use the data:

Futga: Towards Fine-grained Music Understanding through Temporally-enhanced Generative Augmentation
Wu, Junda and Novack, Zachary and Namburi, Amit and Dai, Jiaheng and Dong, Hao-Wen and Xie, Zhouhang and Chen, Carol and McAuley, Julian
arXiv, 2024
pdf


PDMX

Description

We introduce PDMX: a Public Domain MusicXML dataset for symbolic music processing. To our knowledge, PDMX is the largest publicly available, copyright-free MusicXML dataset in existence. PDMX includes genre, tag, description, and popularity metadata for every file.

Basic statistics

Format: MusicXML
Hours: 6,250
Size: 254,077

Metadata

See the Zenodo project page for a more in-depth description of the metadata. However, at an abstract level, the metadata includes:

Example

path,metadata,has_metadata,version,is_user_pro,is_user_publisher,is_us er_staff,has_paywall,is_rated,is_official,is_original,is_draft,has_cus tom_audio,has_custom_video,n_comments,n_favorites,n_views,n_ratings,ra ting,license,license_url,genres,groups,tags,song_name,title,subtitle,a rtist_name,composer_name,publisher,complexity,n_tracks,tracks,song_len gth,song_length.seconds,song_length.bars,song_length.beats,n_notes,not es_per_bar,n_annotations,has_annotations,n_lyrics,has_lyrics,n_tokens, pitch_class_entropy,scale_consistency,groove_consistency,best_path,is_ best_path,best_arrangement,is_best_arrangement,best_unique_arrangement ,is_best_unique_arrangement,subset:all,subset:rated,subset:deduplicate d,subset:rated_deduplicated ./data/b/b/QmbbGKtZ9G6DkWxvSeU516c1ktWiFJmEbHGmR3JFtLAPyC.json,./metad ata/5/5740212.json,True,3.01,True,False,False,False,True,False,False,F alse,False,False,0,2,97,3,4.66,publicdomain,https://creativecommons.or g/publicdomain/mark/1.0/,classical,NA,NA,I shall be no stranger there, I shall be no stranger there - William J. Kirkpatrick,NA,William J. Kirkpatrick,William James Kirkpatrick,NA,0,2,0-0,33120,41.3999171990892,26,71,360,13.846153846153848,4,True,174,True, 538,2.772587471407348,0.9607843137254902,0.8357843137254902,/data2/zac hary/musescore/data/b/b/QmbbGKtZ9G6DkWxvSeU516c1ktWiFJmEbHGmR3JFtLAPyC .mscz,True,/data2/zachary/musescore/data/b/b/QmbbGKtZ9G6DkWxvSeU516c1k tWiFJmEbHGmR3JFtLAPyC.mscz,True,/data2/zachary/musescore/data/b/b/Qmbb GKtZ9G6DkWxvSeU516c1ktWiFJmEbHGmR3JFtLAPyC.mscz,True,True,True,True,Tr ue

Download links

See the Zenodo project page to download PDMX and find more information about the project.

Citation

Please cite the following if you use the data:

PDMX: A Large-Scale Public Domain MusicXML Dataset for Symbolic Music Processing
Phillip Long, Zachary Novack, Taylor Berg-Kirkpatrick, Julian McAuley
arXiv, 2024
pdf


Multi-aspect Reviews

Description

These datasets include reviews with multiple rated dimensions. The most comprehensive of these are beer review datasets from Ratebeer and Beeradvocate, which include sensory aspects such as taste, look, feel, and smell.

Basic statistics

Ratebeer BeerAdvocate
Number of users: 40,213 33,387
Number of items: 110,419 66,051
Number of ratings/reviews: 2,855,232 1,586,259
Timespan: Apr 2000 - Nov 2011 Jan 1998 - Nov 2011

Metadata

Example (ratebeer)

beer/name: John Harvards Simcoe IPA beer/beerId: 63836 beer/brewerId: 8481 beer/ABV: 5.4 beer/style: India Pale Ale (IPA) review/appearance: 4/5 review/aroma: 6/10 review/palate: 3/5 review/taste: 6/10 review/overall: 13/20 review/time: 1157587200 review/profileName: hopdog review/text: On tap at the Springfield, PA location. Poured a deep and cloudy orange (almost a copper) color with a small sized off white head. Aromas or oranges and all around citric. Tastes of oranges, light caramel and a very light grapefruit finish. I too would not believe the 80+ IBUs - I found this one to have a very light bitterness with a medium sweetness to it. Light lacing left on the glass.

Download links

BeerAdvocate (433mb)

RateBeer (388mb)

Sentences with aspect labels (annotator 1) (758kb)

Sentences with aspect labels (annotator 2) (759kb)

Citation

Please cite the following if you use the data:

Learning attitudes and attributes from multi-aspect reviews
Julian McAuley, Jure Leskovec, Dan Jurafsky
International Conference on Data Mining (ICDM), 2012
pdf

From amateurs to connoisseurs: modeling the evolution of user expertise through online reviews
Julian McAuley, Jure Leskovec
WWW, 2013
pdf


Social Circles

Description

These datasets contain social connections and "circles" from Facebook, Twitter, and Google Plus.

Basic statistics

Facebook Twitter Google Plus
Number of networks: 10 133 1,000
Number of nodes: 4,039 106,674 192,075
Number of circles: 193 479 5,541

Metadata

Example (Kaggle egonet data)

UserId: Friends 1: 4 6 12 2 208 2: 5 3 17 90 7

Download links

See SNAP facebook, twitter, and Google Plus data, as well as the Kaggle competition based on the same data.

Citation

Please cite the following if you use the data:

Learning to Discover Social Circles in Ego Networks
Julian McAuley, Jure Leskovec
Neural Information Processing Systems (NIPS), 2012
pdf


Reddit Submissions

Description

Submissions of reddit posts (and in particular resubmissions of the same content) along with metadata.

Basic statistics

Num of submissions (images): 132,308
Num of unique images: 16,736
Timespan July 2008 - January 2013

Metadata

Example

#image_id,unixtime,rawtime,title,total_votes,reddit_id,... number_of_downvotes,localtime,score,number_of_comments,username 1005,1335861624,2012-05-01T15:40:24.968266-07:00,I immediately regret this decision,27,t296r,20,pics,7,1335886824,13,0,ninjaroflmaster 1005,1336470481,2012-05-08T16:48:01.418140-07:00,"Pushing your friend into the water,Level: 99",18,tds4i,16,funny,2,1336495681,14,0,hme4 1005,1339566752,2012-06-13T12:52:32.371941-07:00,I told him. He Didn't Listen,6,v0cma,4,funny,2,1339591952,2,0,HeyPatWhatsUp 1005,1342200476,2012-07-14T00:27:56.857805-07:00,Don't end up as this guy.,16,wjivx,7,funny,9,1342225676,-2,2,catalyst24

Download links

resubmissions data (7.3mb)

raw html of resubmissions (1.8gb)

See also the SNAP project page.

Citation

Please cite the following if you use the data:

Understanding the interplay between titles, content, and communities in social media
Himabindu Lakkaraju, Julian McAuley, Jure Leskovec
ICWSM, 2013
pdf


Questions and comments to Julian McAuley