Turkish EV Charging Intent Dataset
Description
Turkish EV Charging Intent Dataset is a balanced Turkish-language benchmark containing 192 synthetic-editorial electric-vehicle and charging queries across eight intent classes, with fixed train (128), validation (32), and test (32) splits. It is intended for Turkish intent classification, routing evaluation, and deterministic regression testing. The dataset contains no search logs, customer conversations, or personal data. This Mendeley Data record is a distribution mirror of version 1.0.0. Canonical citation and permanent archive: https://doi.org/10.5281/zenodo.22062688. Documentation: https://tekpedal.com/acik-veri. Source repository: https://github.com/gokimedia/turkish-ev-charging-intents
Files
Steps to reproduce
1. Download Turkish-EV-Charging-Intent-Dataset-v1.0.0.zip. 2. Verify SHA-256: 4F60CFE6C0802ADFC8A8DBD226D127151ABFAC133753935966A972BCA5583C00. 3. Extract the archive and run npm test to validate 192 rows, eight balanced intents, and the fixed train/validation/test splits. 4. Load the JSONL splits directly or use turkish-ev-charging-intents.csv for tabular workflows. 5. Keep the provided test split unchanged when reporting comparable benchmark results.