# Dataset Card for "stenotype-eval-js"

This is one of the datasets used to evaluate StenoType ([model](https://huggingface.co/nuprl/stenotype), [GitHub](https://github.com/nuprl/StenoType/)).
This dataset is called `typeweaver-bundle-filtered-subset` in the code and `JS-Sourced` in the dissertation.

The dataset is derived from the [TypeWeaver benchmarks](https://github.com/nuprl/TypeWeaver/tree/main/data/original), with multi-file projects bundled
into single files, and filtered. For more details, see the StenoType GitHub repository and the linked dissertation.

To type check this dataset, a tarball containing the type declaration (`.d.ts`) files is included in `type_declarations.tar.gz`.

## Versions

The evaluation dataset is a sample of 50 files from the "filtered" dataset, which can be found in the branch `filtered`.

The filtered dataset is derived from the full dataset, which can be found in the branch `full`.