task_categories:
- question-answering
language:
- en
tags:
- not-for-all-audiences
- alpaca
pretty_name: Retro Yahoo! Answers
size_categories:
- 1M<n<10M
Description
This dataset is an instruct style dataset comprised of a scrape of the Yahoo! Answers website that was done in 2007. The dataset is comprised of 10 categories labeled 1-10. The categories are as follows:
- Society & Culture
- Science & Mathematics
- Health
- Education & Reference
- Computers & Internet
- Sports
- Business & Finance
- Entertainment & Music
- Family & Relationships
- Politics & Government
The subject line and body of the question have been combined into a single field and separated by a newline character.
I would caution against using this dataset for any serious application as it contains hilariously out of date information, offensive language, and frequent spelling and grammar errors. It is, however, a charming snapshot of the internet in 2007.
Roughly 228m llama tokens in 1.4m samples
Original README
Yahoo! Answers Topic Classification Dataset
Version 2, Updated 09/09/2015
ORIGIN
The original Yahoo! Answers corpus can be obtained through the Yahoo! Research Alliance Webscope program. The dataset is to be used for approved non-commercial research purposes by recipients who have signed a Data Sharing Agreement with Yahoo!. The dataset is the Yahoo! Answers corpus as of 10/25/2007. It includes all the questions and their corresponding answers. The corpus contains 4483032 questions and their answers.
The Yahoo! Answers topic classification dataset is constructed by Xiang Zhang ([email protected]) from the above dataset. It is used as a text classification benchmark in the following paper: Xiang Zhang, Junbo Zhao, Yann LeCun. Character-level Convolutional Networks for Text Classification. Advances in Neural Information Processing Systems 28 (NIPS 2015).
DESCRIPTION
The Yahoo! Answers topic classification dataset is constructed using 10 largest main categories. Each class contains 140,000 training samples and 6,000 testing samples. Therefore, the total number of training samples is 1,400,000 and testing samples 60,000 in this dataset. From all the answers and other meta-information, we only used the best answer content and the main category information.
The file classes.txt contains a list of classes corresponding to each label.
The files train.csv and test.csv contain all the training samples as comma-sparated values. There are 4 columns in them, corresponding to class index (1 to 10), question title, question content and best answer. The text fields are escaped using double quotes ("), and any internal double quote is escaped by 2 double quotes (""). New lines are escaped by a backslash followed with an "n" character, that is "\n".