Skip to content

Dataset · ibm-research

Movie plot question and answer dataset

ibm-research/duorc

A corpus of 100K to 1M English QA pairs based on Wikipedia and IMDb movie descriptions.

What's inside

The dataset contains English questions and answers collected by Amazon Mechanical Turk crowdsourcers based on Wikipedia and IMDb movie plots, where authors could phrase answers in their own words. It contains between 100,000 and 1 million records divided into two parts: SelfRC (based on Wikipedia texts only) and ParaphraseRC (questions based on Wikipedia, but answers based on IMDb plot versions).

What you can do with it

  • Training QA systems to understand long narrative texts and paraphrases
  • Evaluating LLM ability to find non-obvious facts across different versions of the same plot
Tasks
Question answering
Languages
English
Size
100K–1M rows
Downloads this month
3.4K
Likes
36

How to load it

from datasets import load_dataset

ds = load_dataset("ibm-research/duorc")