ADReSS
An authoritative benchmark dataset in the field of Alzheimer's disease speech detection, used to evaluate automatic detection methods for cognitive impairment based on speech analysis.
Timeline (last 90 days)
The ADReSS test set is small in scale and homogeneous in population composition, so improvements on this benchmark may not directly generalize to real clinical populations.
The ADReSS benchmark is a public challenge dataset launched at INTERSPEECH 2020, containing Cookie Theft picture-description recordings and transcripts from 78 AD patients and 78 healthy controls, with the official training/test sets split evenly.
The evaluation metrics of the ADReSS benchmark are accuracy and F1, and in recent years most systems' accuracy has clustered in the 75%–85% range.
NarraSpace's evaluation is based on the 48-person test set of the ADReSS benchmark, so the sample size is relatively limited, and the coverage of predefined scene events may also be constrained by the design of the image itself.
NarraSpace achieves an average accuracy of 80.88%±2.56% across 50 random seeds on the official 48-person test set of the ADReSS benchmark.
After averaging probabilities, NarraSpace achieves 81.25% accuracy and an AUC of 91.32% on the ADReSS test set.
Speaker-independent experiments on ADReSS and ADReSSo show that dual-task vector merging outperforms frozen probing, LoRA, and full fine-tuning.
ADReSS and ADReSSo are INTERSPEECH series challenge datasets, both binary classification (AD vs. healthy control), and emphasize speaker-independent splits.
The research team evaluated LAPE on two standard datasets, ADReSS and ADReSSo.
All Facts (9)
The ADReSS test set is small in scale and homogeneous in population composition, so improvements on this benchmark may not directly generalize to real clinical populations.
50%UnverifiedThe ADReSS benchmark is a public challenge dataset launched at INTERSPEECH 2020, containing Cookie Theft picture-description recordings and transcripts from 78 AD patients and 78 healthy controls, with the official training/test sets split evenly.
50%UnverifiedThe evaluation metrics of the ADReSS benchmark are accuracy and F1, and in recent years most systems' accuracy has clustered in the 75%–85% range.
50%UnverifiedNarraSpace's evaluation is based on the 48-person test set of the ADReSS benchmark, so the sample size is relatively limited, and the coverage of predefined scene events may also be constrained by the design of the image itself.
50%UnverifiedNarraSpace achieves an average accuracy of 80.88%±2.56% across 50 random seeds on the official 48-person test set of the ADReSS benchmark.
50%UnverifiedAfter averaging probabilities, NarraSpace achieves 81.25% accuracy and an AUC of 91.32% on the ADReSS test set.
50%UnverifiedSpeaker-independent experiments on ADReSS and ADReSSo show that dual-task vector merging outperforms frozen probing, LoRA, and full fine-tuning.
50%UnverifiedADReSS and ADReSSo are INTERSPEECH series challenge datasets, both binary classification (AD vs. healthy control), and emphasize speaker-independent splits.
50%UnverifiedThe research team evaluated LAPE on two standard datasets, ADReSS and ADReSSo.
50%