TurkuNLP/genre-6

Dataset Card for Dataset Name

Dataset Summary

Genre-6 dataset is an English dataset based on Kindletrends (UK & US). It contains more than 20k books and associated categories with ready-made binary classification and multilabel classification labels.

Dataset Structure

Data Instances

{"text": "...", "categories": "Engineering & Transportation;Science & Math", "fiction": "non-fiction", "split1": ['Science & Math'], "split2" : ['Engineering & Transportation', 'Science & Math'], "split3": ['Science & Math']}

Data Fields

• text: Kindletrends text
• categories: Kidletrends categories (1 to 2 categories per book)
• fiction: binary label for fiction and non-fiction books
• splits 1,2,3: multilabel for different subsets of the categories

Data Splits

The dataset contains train (80%), validation (10%) and test (10%) splits.

The splits for multilabels are following:

• split1: 'Biology & Nature & Biological Sciences','Computer Science', 'Fantasy','Medicine & Health Sciences','Philosophy','Science & Math'.
• split2: 'Biology & Nature & Biological Sciences','Computer Science', 'Engineering & Transportation','Fantasy','Medicine & Health Sciences','Science & Math'.
• split3: 'Biology & Nature & Biological Sciences','Computer Science', 'Fantasy','Medicine & Health Sciences', 'Poetry', 'Politics & Social Sciences', 'Science & Math'.

More splits can be generated from the field "categories".

Source Data

Kindletrends