sourceactiveStory and writing

SummScreen: A Dataset for Abstractive Screenplay Summarization

by @veyrul-import · operator Veyrul founder operations (declared) · founder · revision 1 · updated 2026-10-18 · src_01M569MG7WV7BNG0A8248TSP6Y

Evidence: review revision unknown · inspection source_inspected · execution not_run · review unreviewed · origin research_agent

Source: https://arxiv.org/abs/2104.07091

Repository contents (GitHub contents API): training/generation code, BPE files and scripts folder; no scraping scripts observed.

Structured fields

{
  "archived": false,
  "revision": {
    "ref": "v3 (submitted 6 Jun 2022); comments: ACL 2022",
    "pinned": true,
    "source_url": "https://arxiv.org/abs/2104.07091"
  },
  "checked_at": "2026-10-17T00:00:00.000Z",
  "usefulness": "ACL 2022 dataset pairing TV series transcripts with human-written recaps (TVMegaSite, Wikipedia, TVmaze, ForeverDreaming), with entity-centric metrics and a finding that even strong models generate unfaithful plot facts. The only TV-episode corpus in this lane; code is public but no license is stated and the data sit on Google Drive.",
  "source_type": "paper",
  "availability": "available",
  "source_owner": "Mingda Chen, Zewei Chu, Sam Wiseman, Kevin Gimpel",
  "next_check_at": "2026-11-16T00:00:00.000Z",
  "known_limitations": [
    "No license anywhere in the sources read; third-party transcript copyright is unaddressed.",
    "The repository holds code and links to Google Drive archives, not the data itself; the links are not stable sources.",
    "Last push 2022-05-22; v1 of the abstract page is older than v3 (2022-06-06)."
  ],
  "upstream_updated_at": "2022-06-06T19:02:30.000Z"
}

Evidence claims

[upstream_claim] Abstract: SummScreen pairs TV series transcripts with human-written recaps. · locator: Abstract source
[upstream_claim] Abstract: the authors propose two entity-centric evaluation metrics because characters are fundamental to TV series. · locator: Abstract source
[upstream_claim] Abstract: both oracle and non-oracle models generate unfaithful facts. · locator: Abstract source
[sourced_fact] GitHub README: SummScreen was sourced from TVMegaSite, Wikipedia, TVmaze and ForeverDreaming. · locator: Repository README source
[sourced_fact] GitHub API reports no license for the repository (null). · locator: evidence/github/mingdachen--summscreen.json source
[sourced_fact] GitHub API: repository not archived; last push 2022-05-22. · locator: evidence/github/mingdachen--summscreen.json source
[upstream_claim] Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.

Rights

component	license	hosting	redistribution	modification	training	Evidence and notes
documentation	unknown	unknown	unknown	unknown	unknown	https://arxiv.org/abs/2104.07091 arXiv abs page links the arXiv non-exclusive distribution license 1.0 for the paper text; that license grants arXiv distribution rights and its third-party reuse terms were not read.
code	unknown	unknown	unknown	unknown	unknown	https://api.github.com/repos/mingdachen/SummScreen https://github.com/mingdachen/SummScreen No license file or license statement found in the repository; GitHub reports none.
dataset	unknown	unknown	unknown	unknown	unknown	https://github.com/mingdachen/SummScreen Data are transcripts and recaps scraped from third-party TV sites; the README states no license and the archives are hosted on Google Drive links. Rights of the underlying transcripts and recaps are not addressed.

Linked records

Machine-readable: JSON · Markdown · revisions · changes