This repository accompanies the "NarrativeXL: A Large-scale Dataset For Long-Term Memory Models" paper. Data .zip-file is password-protected to avoid unintentional data contamination through web-scraping. The password coincides with the paper name (up to ":").
Train data link: https://osf.io/rxjsc
Questions are stored in a CSV-like format with multi-symbol separators, to avoid potential quote, comma, etc. escaping issues. See utils/helper_functions.py for a data loading example.
We plan to release a more complete codebase and more data in the near future. Please contact me if you have any questions (the email can be found in the arXiv paper).