Dispatch · Folklore

New Software Tool Extracts Giants, Spirits and Shapeshifters from Myth Texts

· via exa.ai

New Software Tool Extracts Giants, Spirits and Shapeshifters from Myth Texts

Jason Meade released a software resource on Zenodo (CERN) on September 24, 2026, titled “Folklore Entity Extraction Pipeline.” According to its description, the pipeline reads mythological, ethnographic, or folkloric text files and extracts structured records of non-human humanoid entities, including giants, dwarfs, wild men, monsters, spirits, shapeshifters, and animal-humanoids. It uses the DeepSeek API with a configurable external prompt and writes JSONL output, one entity per line, with each entity traced back to its source file and chunk. The package also contains scripts for translating entities, normalizing traits, and extracting and cleaning citations, along with a taxonomy file and documentation.

The author describes the tool as resumable and cost-scalable, and estimates that a full run on a 1,000-file corpus costs $20 to $50. It requires Python 3.8 or later, the requests package, and a DeepSeek API key. Meade developed it for a comparative mythology project and released it as a standalone tool for researchers working with large corpora of mythological or folkloric text. For readers who follow giants, wild men, and shapeshifters across traditions, the practical value lies in traceability: every record points back to the passage it came from. The release is a method, not a set of conclusions about the beings themselves.

Common Questions

What does this pipeline actually do?
According to its description, it reads mythological, ethnographic, or folkloric text files and pulls out structured records of non-human humanoid entities such as giants, dwarfs, wild men, spirits, and shapeshifters. The output is a JSONL file with one entity per line, each traced to its source file and chunk.
Does it reveal anything new about giants or spirits in myth?
No. The release is a software tool developed for a comparative mythology project, and the author does not present conclusions about the entities themselves. It is offered so researchers can process large folkloric corpora in a structured, traceable way.