# Family Fail survey credits

Family Fail uses 9,672 sourced questions adapted from the **ProtoQA dataset** by the University of Massachusetts Amherst IESL research group. Dataset: https://github.com/iesl/protoqa-data . Paper: https://arxiv.org/abs/2005.00771 . Licensed under Creative Commons Attribution 4.0; the complete license is preserved in [PROTOQA-LICENSE.txt](PROTOQA-LICENSE.txt).

The training and scraped development sets contain questions and answer clusters collected from Family Feud fan sites. The original source URL and dataset identifier are retained with every server-side record. These are sourced historical answer counts, not surveys conducted by GandyStudios. No simulated surveys were added.

Changes: whitespace cleanup, ordering by answer counts, duplicate-question removal, and excluding entries with unsupported answer counts or lengths. Answers and unrevealed point values stay on the server during play. Family Fail is an independent parody game and is not affiliated with the television show.