{
  "@context": {
    "@language": "en",
    "@vocab": "https://schema.org/",
    "citeAs": "cr:citeAs",
    "column": "cr:column",
    "conformsTo": "dct:conformsTo",
    "cr": "http://mlcommons.org/croissant/",
    "rai": "http://mlcommons.org/croissant/RAI/",
    "data": {
      "@id": "cr:data",
      "@type": "@json"
    },
    "dataType": {
      "@id": "cr:dataType",
      "@type": "@vocab"
    },
    "dct": "http://purl.org/dc/terms/",
    "prov": "http://www.w3.org/ns/prov#",
    "examples": {
      "@id": "cr:examples",
      "@type": "@json"
    },
    "extract": "cr:extract",
    "field": "cr:field",
    "fileProperty": "cr:fileProperty",
    "fileObject": "cr:fileObject",
    "fileSet": "cr:fileSet",
    "format": "cr:format",
    "includes": "cr:includes",
    "isLiveDataset": "cr:isLiveDataset",
    "jsonPath": "cr:jsonPath",
    "key": "cr:key",
    "md5": "cr:md5",
    "parentField": "cr:parentField",
    "path": "cr:path",
    "recordSet": "cr:recordSet",
    "references": "cr:references",
    "regex": "cr:regex",
    "repeated": "cr:repeated",
    "replace": "cr:replace",
    "sc": "https://schema.org/",
    "separator": "cr:separator",
    "source": "cr:source",
    "subField": "cr:subField",
    "transform": "cr:transform"
  },
  "@type": "sc:Dataset",
  "conformsTo": [
    "http://mlcommons.org/croissant/1.0",
    "http://mlcommons.org/croissant/RAI/1.0"
  ],
  "name": "Classical Hollywood Narrative Benchmark",
  "description": "The Classical Hollywood Narrative Benchmark (CHNB) is a multiple-choice question benchmark for long-form video understanding of narrative.  Each question has four answer choices (A, B, C, D) and only one correct anwer. Questions are keyed to a movie_id and the movie_id is linked to a URL of an mp4 in movies.jsonl",
  "url": "https://github.com/bamman-group/chnb",
  "version": "1.0.0",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "datePublished": "2026",  
  "citeAs": "David Bamman, Kent Chang, Allison Cooper, Juishan Hsu, Reina Kushihashi, Madison Mar, Arnav Podichetty, Rachael Samberg, Ipek Nil Sancak, and Yuhan Shao (2026). Evaluating Multimodal Narrative Understanding of Popular Hollywood Films. ArXiV (forthcoming).",
  "creator": {
    "@type": "Person",
    "name": "David Bamman",
    "email": "dbamman@berkeley.edu",
    "affiliation": {
      "@type": "Organization",
      "name": "University of California, Berkeley"
    }
  },
  "keywords": [
    "multimodal question answering",
    "narrative understanding",
    "long-form video"
  ],
  "rai:dataCollection": "Movies were sourced through three requirements: popularity (among the top 100 movies in a year, as measured in box office earnings reported through Variety magazine); public domain status (either movies published before 1931 or movies published in 1931 and afterward that registered their copyright and where no record of renewal can be found); and with an mp4 file currently available on the Internet Archive. For each movie, trained annotators wrote multiple-choice questions targeting temporality, plot, character, setting, perspective, representation, symbolism and object identification; each question has one correct answer along with three plausible distractors.",
  "rai:dataCollectionType": [
    "Secondary Data Analysis",
    "Manual Human Curator"
  ],
  "rai:dataCollectionRawData": "The raw data consists of feature-length movie video files (.mp4) originally hosted on the Internet Archive. The questions, answer choices, and gold labels were authored from scratch by human annotators while watching the films.",
  "rai:dataCollectionMissingData": "No data is missing.",
  "rai:dataCollectionTimeframe": [
    {
      "@value": "2026-01-01T00:00:00",
      "@type": "sc:DateTime"
    },
    {
      "@value": "2026-04-30T00:00:00",
      "@type": "sc:DateTime"
    }
  ],

  "rai:dataManipulationProtocol": "The underlying movie content was not altered. Question and answer options were reviewed for surface-level spelling, grammar and logic.",
  "rai:dataAnnotationProtocol": "Each question was authored by one annotator. Annotators received training covering question types, distractor quality, and disallowed content (e.g., questions answerable from the movie title alone).",
  "rai:dataAnnotationPlatform": [
    "Annotation was performed by watching the movie and writing questions and answer options in a spreadsheet."
  ],
  "rai:dataAnnotationAnalysis": [
    "At the end of annotation, all co-authors carried out a validity check by independently watching one movie they were each unfamiliar with and answering all questions for that movie that a different annotator had written."
  ],
  "rai:annotatorDemographics": [
    "Annotators were adult fluent English speakers based in the United States, recruited from the project team. No further demographic information is collected or released."
  ],
  "rai:machineAnnotationTools": [
    "No machine annotation tools, including LLMs or AI, were used in creating benchmark questions. Annotators made use of LLMs (in particular, an institutional Gemini subscription whose terms of use prevent Google from training models on user chats) to assess whether a question is answerable from the movie title and date alone. LLMs were not used to generate the questions or answers themselves."
  ],
  "rai:hasSyntheticData": false,
  "prov:wasGeneratedBy": {
    "@type": "prov:Activity",
    "prov:label": "Classical Hollywood Narrative Benchmark construction",
    "description": "Construction of the Classical Hollywood Narrative Benchmark, comprising (1) curation of a catalog of feature-length movies and their .mp4 URLs and (2) human authoring of multiple-choice questions and distractors while watching each film",
    "prov:startedAtTime": {
      "@value": "2026-01-01T00:00:00",
      "@type": "sc:DateTime"
    },
    "prov:endedAtTime": {
      "@value": "2026-04-30T00:00:00",
      "@type": "sc:DateTime"
    },
    "prov:wasAssociatedWith": {
      "@type": "prov:Agent",
      "name": "David Bamman",
      "email": "dbamman@berkeley.edu",
      "affiliation": {
        "@type": "Organization",
        "name": "University of California, Berkeley"
      }
    }
  },
  "rai:annotationsPerItem": "Each question was written by one annotator.",
  "rai:personalSensitiveInformation": [
    "The dataset contains popular, publicly released movies, along with questions writen about those movies. No additional personal data about annotators or other entities not found in those films is included."
  ],
  "rai:dataBiases": [
    "Movies in the dataset are selected based on popularity in the United States (as reported through Variety magazine) and reflect English-language cinematic traditions.",
    "Questions were authored by annotators at UC Berkeley, which may bias what is treated as salient or noteworthy in a film."
  ],
  "rai:dataUseCases": [
    "Benchmarking long-form video understanding, multimodal reasoning, and narrative question answering over movies (testing/validation)"
  ],
  "rai:dataLimitations": [
    "Coverage is limited to the catalog of popular movies that are likely in the public domain and the question types written by the annotation team.",
    "Movies were all originally released prior to 1964 and may not reflect later cinematic techniques.",
    "The films that comprise this benchmark were selected based on publication years prior to 1931 or to lack of evidence of required registration renewal; while we exclude several categories of protected films (e.g., adaptations of copyrighted works), inclusion does not constitute a legal warranty of public domain status."
  ],
  "rai:dataSocialImpact": "The Classical Hollywood Narrative Benchmark is released as an open benchmark of popular films that are likely in the public domain, in order to drive future research in computer vision, AI and cultural analytics on a stable collection of films (those with minimal risk of disappearing through claims of copyright infringement). Potential negative impacts include reinforcing the cultural biases of the underlying film catalog.",
  "distribution": [
    {
      "@type": "cr:FileObject",
      "@id": "github-repository",
      "name": "github-repository",
      "description": "Classical Hollywood Narrative Benchmark release on GitHub.",
      "contentUrl": "https://github.com/bamman-group/chnb",
      "encodingFormat": "git+https",
      "sha256": "main"
    },
    {
      "@type": "cr:FileObject",
      "@id": "benchmark-jsonl",
      "name": "benchmark.jsonl",
      "description": "JSON Lines file containing the benchmark multiple-choice questions. One question per line.",
      "containedIn": {
        "@id": "github-repository"
      },
      "contentUrl": "data/benchmark.jsonl",
      "encodingFormat": "application/jsonlines"
    },
    {
      "@type": "cr:FileObject",
      "@id": "non_benchmark-jsonl",
      "name": "non_benchmark.jsonl",
      "description": "JSON Lines file containing supplemental non-benchmark multiple-choice questions. One question per line.",
      "containedIn": {
        "@id": "github-repository"
      },
      "contentUrl": "data/non_benchmark.jsonl",
      "encodingFormat": "application/jsonlines"
    },
    {
      "@type": "cr:FileObject",
      "@id": "movie-metadata.tsv",
      "name": "movie_metadata.tsv",
      "description": "TSV file with metadata for each movie, including URL for mp4. One movie per line.",
      "containedIn": {
        "@id": "github-repository"
      },
      "contentUrl": "data/movie_metadata.tsv",
      "encodingFormat": "text/tab-separated-values"
    }
  ],
  "recordSet": [
    {
      "@type": "cr:RecordSet",
      "@id": "movies",
      "name": "movies",
      "description": "Catalog of movies referenced by the benchmark, each with a downloadable .mp4 URL.",
      "key": {
        "@id": "movies/video_file_id"
      },
      "field": [
        {
          "@type": "cr:Field",
          "@id": "movies/video_file_id",
          "name": "video_file_id",
          "description": "Unique identifier for the movie.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "video_file_id"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/title",
          "name": "title",
          "description": "Title of the movie.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "title"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/year",
          "name": "year",
          "description": "Release year of the movie.",
          "dataType": "sc:Integer",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "year"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/IMDB",
          "name": "IMDB",
          "description": "IMDb identifier.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "IMDB"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/genre",
          "name": "genre",
          "description": "Comma-separated list of genres.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "genre"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/director",
          "name": "director",
          "description": "Director of the movie.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "director"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "movies/IA_mp4_url",
          "name": "IA_mp4_url",
          "description": "Direct URL to the movie .mp4 file on the Internet Archive.",
          "dataType": [
            "sc:URL",
            "sc:VideoObject"
          ],
          "encodingFormat": "video/mp4",
          "source": {
            "fileObject": {
              "@id": "movie-metadata.tsv"
            },
            "extract": {
              "column": "IA_mp4_url"
            }
          }
        }
      ]
    },
    {
      "@type": "cr:RecordSet",
      "@id": "questions",
      "name": "questions",
      "description": "Multiple-choice questions about movies. Each row has four answer choices (A-D) and exactly one correct answer.",
      "key": {
        "@id": "questions/qid"
      },
      "field": [
        {
          "@type": "cr:Field",
          "@id": "questions/qid",
          "name": "qid",
          "description": "Unique identifier for the question.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "qid"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/video_file_id",
          "name": "video_file_id",
          "description": "Key into the movies record set; identifies the movie this question is about.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "video_file_id"
            }
          },
          "references": {
            "field": {
              "@id": "movies/video_file_id"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/category",
          "name": "category",
          "description": "Question category (e.g., Temporality, Character, Plot).",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "category"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/question",
          "name": "question",
          "description": "The question text.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "question"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/option_A",
          "name": "option_A",
          "description": "Answer choice A.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "option_A"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/option_B",
          "name": "option_B",
          "description": "Answer choice B.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "option_B"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/option_C",
          "name": "option_C",
          "description": "Answer choice C.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "option_C"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/option_D",
          "name": "option_D",
          "description": "Answer choice D.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "option_D"
            }
          }
        },
        {
          "@type": "cr:Field",
          "@id": "questions/answer",
          "name": "answer",
          "description": "The letter (A, B, C, or D) of the correct answer choice.",
          "dataType": "sc:Text",
          "source": {
            "fileObject": {
              "@id": "benchmark-jsonl"
            },
            "extract": {
              "column": "answer"
            }
          }
        }
      ]
    }
  ]
}