arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2604.06758cs.CL
收藏

基于基础模型时代的多语言认知障碍检测

Multilingual Cognitive Impairment Detection in the Era of Foundation Models

Jožef Stefan Institute(约瑟夫·斯特凡研究所);Jožef Stefan International Postgraduate School(约瑟夫·斯特凡国际研究生院);Institute of Synergetics and Psychotherapy Research, Paracelsus Medical University(协同与心理治疗研究所,帕拉塞尔苏斯医科大学);University of Ljubljana(卢布尔雅那大学);Queen Mary University of London(伦敦玛丽女王大学)

Damar Hoogland, Boshko Koloski, Jaya Caporusso, Tine Kolenik, Ana Zwitter Vitez, Senja Pollak, Christina Manouilidou, Matthew Purver

AI总结:

本文比较了零样本大语言模型与监督表格方法在多语言语音转录中的认知障碍分类性能,发现监督模型在包含工程语言特征时表现更优。

摘要:

我们评估了从英语、斯洛文尼亚语和韩语的语音转录中进行认知障碍(CI)分类。我们比较了三种输入设置下作为直接分类器使用的零样本大语言模型(LLMs)与监督表格方法,后者在留一法协议下进行训练。表格模型处理工程语言特征、转录嵌入和两种模态的早期或晚期融合。在不同语言中,零样本LLMs提供了有竞争力的无训练基线,但监督表格模型总体表现更好,特别是在包含工程语言特征并结合嵌入时。聚焦于嵌入的少量实验表明,有限监督的价值具有语言依赖性,某些语言显著受益于额外标注示例,而其他语言在没有更丰富的特征表示时仍受限制。总体而言,结果表明,在小数据CI检测中,结构化语言信号和简单的融合基分类器仍保持强大和可靠的信号。

英文摘要:

We evaluate cognitive impairment (CI) classification from transcripts of speech in English, Slovene, and Korean. We compare zero-shot large language models (LLMs) used as direct classifiers under three input settings -- transcript-only, linguistic-features-only, and combined -- with supervised tabular approaches trained under a leave-one-out protocol. The tabular models operate on engineered linguistic features, transcript embeddings, and early or late fusion of both modalities. Across languages, zero-shot LLMs provide competitive no-training baselines, but supervised tabular models generally perform better, particularly when engineered linguistic features are included and combined with embeddings. Few-shot experiments focusing on embeddings indicate that the value of limited supervision is language-dependent, with some languages benefiting substantially from additional labelled examples while others remain constrained without richer feature representations. Overall, the results suggest that, in small-data CI detection, structured linguistic signals and simple fusion-based classifiers remain strong and reliable signals.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑