censor-board-cuts

Note: Data updates paused as of June 12th, due to breaking changes on the e-Cinepramaan portal

Dataset and related anlysis of modifications or cuts made by the Central Board of Film Certification (CBFC), India.

The dataset consists of two main components:

• Raw Data: Raw category and certificate data from the CBFC website, stored in data/raw/
• Processed Data: Cleaned up data enhanced with code-based and LLM-based analysis of cuts, stored in data/data.csv

Preview

• Modifications (~20MB)
• Metadata (~40MB)
• Categories (~100MB)
• Processed Dataset (~100MB)

Further data is available in the data/ directory.

Data Collection

The following scripts fetch data from the CBFC website:

• scripts/certificates/: Film metadata, modifications
• scripts/categories/: Film categories

The above scripts incrementally fetch new films and append them to the relevant CSV files. After fetching the data from the CBFC website, code-based analysis of the metadata and modifications is done in scripts/analysis/ and LLM-based analysis is done in scripts/llm/. Next, scripts/imdb/ further enhances the metadata and all the fetched data is joined together using scripts/join/ which saves the final data in data/data.csv.

Data Analysis

The code-based analysis is done by a Python script scripts/analysis/main.py that cleans and processes the raw data:

• Standardizes duration formats and attempts to pull out timestamps from the descriptions.
• Categorizes modifications based on type (audio, visual, deletion, etc.) and the basic type of content (violence, nudity, etc.) using an LLM.

TODO

• Create a dashboard for exploring the data.

Related Projects

• https://github.com/harman28/cbfc
• https://github.com/ananddotiyer/MovieCertifications