AIeathumberger/Aphasiabench

AphasiaBench

<details> <summary><strong>中文 (点击展开)</strong></summary>

AphasiaBench 数据集完整版的学术许可证是免费的，可供满足以下条件的研究人员使用：

• —您是学术机构的学生、教授或研究助理。
• —您的研究主要不是为了商业第三方的利益。
• —您正在构建一个仅用于学术目的的工具。
• —在发布研究结果时，您应该承认使用了 AphasiaSimplified 数据资源。

</details>

<details> <summary><strong>English (Click to expand)</strong></summary>

The full version of the AphasiaBench dataset is available free of charge under an academic license for researchers who meet the following conditions:

• —You are a student, professor, or research assistant at an academic institution.
• —Your research is not primarily for the benefit of a commercial third party.
• —You are developing a tool that is intended solely for academic purposes.
• —When publishing your research results, you should acknowledge the use of the AphasiaSimplified data resource.

</details>

Dataset Summary

This dataset is generated by a large model and covers eight language scenarios and five aphasia categories.

Prompt Example for CLAP Framework

We additionally provide a representative prompt example used in the CLAP framework for aphasic text generation.

The example illustrates how clinical rules and aphasia-specific constraints are integrated into large language model prompting, including:

• —emergency scenario description,
• —rule-based control of linguistic structures,
• —and Broca’s aphasia style simulation.

Files:

• —brocapromptexample.png : visual illustration of the prompt and its generated output

All contents are synthetic and provided only for research and demonstration purposes.