README.md

metadata

license: cc-by-4.0
language:
  - zh
pretty_name: CSRTE-Text
tags:
  - text-relation-extraction
  - relation-extraction
  - chinese

CSRTE-Text

CSRTE-Text is the text-only counterpart of the CSRTE Chinese relation extraction benchmark. It contains transcriptions with named-entity and relation annotations, and is intended for text-only SpeechRE experiments, analysis, and comparison with the audio-based benchmark.

Data format

The dataset is provided in train.json, dev.json, and test.json. Each JSON record contains:

• sentence: the transcription of the utterance.
• audio_id: the identifier inherited from the paired speech record.
• data_annotation: entity spans and relation annotations. Entity spans use start, end, text, and labels; relations identify their source and target entities and relation labels.
• data_info: available metadata associated with the utterance.

dataset_info.json lists the entity and relation label sets used by CSRTE.

License

The dataset is released under the Creative Commons Attribution 4.0 International license. Please check the licenses and terms of the underlying sources before redistribution or commercial use.

Citation

CSRTE was introduced and used by the following work:

@inproceedings{mu-etal-2026-acre,
  title = "{ACRE}: An Acoustic-Cue and Contextual-Knowledge Reasoning Enhancement Framework for Speech Relation Extraction",
  author = "Mu, Wenxuan and
    Ning, Jinzhong and
    Zhao, Di and
    Zhang, Yijia",
  editor = "Lee, Hung-Yi and
    Dev, Sunipa and
    Li, Jessy and
    Elliott, Desmond",
  booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2026",
  month = oct,
  year = "2026",
  address = "Budapest, Hungary",
  publisher = "Association for Computational Linguistics",
  url = "https://aclanthology.org/2026.findings-emnlp.1244/",
  doi = "10.18653/2026.findings-emnlp.1244",
  pages = "26343--26352",
  ISBN = "979-8-89176-462-0"
}