Screenplay Corpus Tokenized for GPT-2 Fine-Tuning

Approximately 94 million tokens of professionally formatted screenplay text, pre-tokenized for direct use in GPT-2 training pipelines. The corpus was derived from the Movie-Script-Database by Aveek Saha and is provided as tokenized JSON splits. The dataset was created by kazkiryu and was last updated on June 9, 2026.

Use Cases

• Fine-tuning GPT-2 models for screenplay generation based on the pre-tokenized corpus.
• Evaluating language model performance on structured, professionally formatted text.
• Analyzing stylistic patterns in movie scripts using the provided tokenized data.

Strengths

• Contains approximately 94 million tokens of screenplay text.
• Data is pre-tokenized and ready for direct consumption by a GPT-2 Trainer pipeline, requiring no preprocessing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• Last updated 2026-06-09 06:59:41; freshness should be verified.

Provenance

Source

Derived from Movie-Script-Database by Aveek Saha.

Collection Method

Pre-tokenized screenplay corpus.

Freshness

Last updated 2026-06-09 06:59:41.