metadata
dataset_info:
features:
- name: query
dtype: string
- name: book_title
dtype: string
- name: description
dtype: string
- name: query_id
dtype: int64
- name: row_id
dtype: int64
splits:
- name: train
num_bytes: 2840985
num_examples: 2245
download_size: 1720371
dataset_size: 2840985
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
The queries in this dataset were generated by ChatGPT asking for the most popular book genres and then asking to generate queries of max 3 words to search for books by mood or theme about those genres.
Books titles and descriptions were obtained from Google Books API.
Data cleaning:
- Remove descriptions smaller than 200 chars
- Remove queries with less than 7 books