Useful blunders: Can automated speech recognition errors improve downstream dementia classification?
Useful blunders: Can automated speech recognition errors improve downstream dementia classification?
复制标题
有用的错误：自动语音识别错误能否改善下游痴呆症分类？
DOI：
10.1016/j.jbi.2024.104598
复制
发表时间：
2024
影响因子：
4.5
通讯作者：
Pakhomov,Serguei
中科院分区：
文献类型：
--
作者：
Li,Changye;Xu,Weizhe;Cohen,Trevor;Pakhomov,Serguei
目的研究自动语音识别（ASR）系统的错误如何影响痴呆症分类的准确性，特别是在“Cookie盗窃”图片描述任务中。我们的目的是评估不完美的asr生成的转录本是否可以为区分认知健康个体和阿尔茨海默病（AD）患者的语言样本提供有价值的信息。方法利用不同的ASR模型进行实验，并利用后期编辑技术对转录本进行优化。这些不完美的ASR转录本和人工转录的转录本都被用作下游痴呆分类的输入。我们进行了全面的误差分析，以比较模型的性能并评估asr生成的转录本在痴呆分类中的有效性。结果在“Cookie盗窃”任务中，simperfect asr生成的转录本在区分AD患者和非AD患者方面令人惊讶地优于手动转录。这些基于ASR的模型超越了之前最先进的方法，表明ASR错误可能包含与痴呆相关的有价值的线索。ASR和分类模型之间的协同作用提高了痴呆分类的总体准确性。结论不完善的ASR转录本可有效捕捉与痴呆相关的语言异常，提高分类任务的准确性。ASR和分类模型之间的协同作用强调了ASR作为评估认知障碍和相关临床应用的有价值工具的潜力。
ObjectivesWe aimed to investigate how errors from automatic speech recognition (ASR) systems affect dementia classification accuracy, specifically in the “Cookie Theft” picture description task. We aimed to assess whether imperfect ASR-generated transcripts could provide valuable information for distinguishing between language samples from cognitively healthy individuals and those with Alzheimer’s disease (AD).MethodsWe conducted experiments using various ASR models, refining their transcripts with post-editing techniques. Both these imperfect ASR transcripts and manually transcribed ones were used as inputs for the downstream dementia classification. We conducted comprehensive error analysis to compare model performance and assess ASR-generated transcript effectiveness in dementia classification.ResultsImperfect ASR-generated transcripts surprisingly outperformed manual transcription for distinguishing between individuals with AD and those without in the “Cookie Theft” task. These ASR-based models surpassed the previous state-of-the-art approach, indicating that ASR errors may contain valuable cues related to dementia. The synergy between ASR and classification models improved overall accuracy in dementia classification.ConclusionImperfect ASR transcripts effectively capture linguistic anomalies linked to dementia, improving accuracy in classification tasks. This synergy between ASR and classification models underscores ASR’s potential as a valuable tool in assessing cognitive impairment and related clinical applications.