Building an Open Turkish EV Charging Intent Dataset

Turkish electric-vehicle assistants perform multiple tasks, such as locating charging stations, comparing prices, planning routes, assessing battery health, and comparing vehicles. To extract accurate intent information reliably from user queries, researchers built an open Turkish EV Charging Intent Dataset. This dataset includes 192 Turkish queries, evenly distributed across eight intent classes.

It is published under the CC BY 4.0 license, and the training, validation, and test sets are fixed. TekPedal, an EV charging map and vehicle decision platform for Turkey, maintains the dataset. The dataset can be explored interactively, inspected on GitHub, or cited using its DOI (10.5281/zenodo.22062688). Intent routing is crucial because different queries require different tools and freshness guarantees.

The dataset helps by explicitly separating each query into the correct retrieval source, product page, or application workflow, making evaluation easier. Each dataset record contains a unique ID, Turkish query, intent identifier, human-readable label, suggested content route, split, language, and provenance marker. The dataset was created editorially from documented subject lists and templates, ensuring reproducibility and privacy.

However, it does not measure real search demand or user intent frequency. The dataset is useful for prototyping Turkish text classifiers, evaluating embedding-based or zero-shot routers, building deterministic regression suites, and demonstrating data-loading and classification workflows. It is not meant to prove that a model is production-ready.

The dataset is available from TekPedal Open Data and Zenodo (DOI 10.5281/zenodo.22062688). The documentation, datasheet, license, machine-readable files, and validation scripts are available in the GitHub repository.