README.md

metadata

language:
  - hy
task_categories:
  - sentence-similarity
license: cc-by-nc-4.0

Armenian STS Dataset

This dataset is part of ArmBench-TextEmbed (Armenian Embedding Model Benchmark). A 3K subset of the Semantic Textual Similarity benchmark (May, 2021) translated to Armenian. Sentence pairs with similarity scores (0-5) for Armenian embedding evaluation.

Citation

If you use this dataset in your research, please cite our paper:

@inproceedings{navasardyan2026less,
  title={Less is More: Adapting Text Embeddings for Low-Resource Languages with Small Scale Noisy Synthetic Data},
  author={Navasardyan, Zaruhi and Bughdaryan, Spartak and Minasyan, Bagratuni and Davtyan, Hrant},
  booktitle={Proceedings of the Workshop on Low-Resource Language Models (LoResLM) at EACL 2026},
  year={2026},
  url={https://arxiv.org/abs/2603.22290}
}