README.md

metadata

language:
  - en
language_creators:
  - found
license: unknown
multilinguality:
  - monolingual
pretty_name: Blue
size_categories:
  - 10K<n<100K
source_datasets:
  - extended|app_reviews
tags:
  - reviews
  - ratings
  - ordinal
  - text
task_categories:
  - text-classification
task_ids:
  - text-scoring
  - sentiment-scoring

This dataset is a quick-and-dirty benchmark for predicting ratings across different domains and on different rating scales based on text. It pulls in a bunch of rating datasets, takes at most 1000 instances from each and combines them into a big dataset.

Requires the kaggle library to be installed, and kaggle API keys passed through environment variables or in ~/.kaggle/kaggle.json. See the Kaggle docs.