Speaker-Aware Temporal Aggregation Strategies on Segment Representations for Depression Detection in Dyadic Interaction: A Benchmark Study

Hugging Face Daily Papers Papers
depression-detection temporal-aggregation speech-analysis benchmarking clinical-speech self-supervised-learning dyadic-interaction

Summary

This paper introduces DEPOOL, a controlled benchmark evaluating six temporal aggregation architectures across six frozen speech backbones for depression detection in dyadic interactions, finding that many configurations collapse into single-class predictions and that robustness should be a key criterion.

Speech-based depression detection compresses features from short audio segments into one speaker-level decision, a step called temporal aggregation rarely studied on its own. Most benchmarks fix a single self-supervised encoder and a single hand-picked layer, so a reported gain may reflect the pipeline rather than the aggregation method itself. We introduce DEPOOL, a controlled benchmark that compares six aggregation architectures with six frozen speech backbones on an English and a Mandarin depression corpus, where each configuration learns which backbone layers matter rather than fixing one by hand. Across the resulting 72-configuration grid, a third of configurations collapse into predicting a single class for every speaker, a failure tied to the backbone as much as to the method, and the architecture that is most stable in a single-seed run becomes unreliable when training repeats across seeds. Robustness to backbone and seed, rather than average accuracy across a single pipeline, should be a first-class benchmarking criterion for temporal aggregation in clinical speech.
Original Article
View Cached Full Text

Cached at: 07/07/26, 06:45 PM

Paper page - Speaker-Aware Temporal Aggregation Strategies on Segment Representations for Depression Detection in Dyadic Interaction: A Benchmark Study

Source: https://huggingface.co/papers/2607.02904

Abstract

Temporal aggregation methods for speech-based depression detection show inconsistent performance across different backbones and training runs, highlighting the need for robust benchmarking criteria.

Speech-baseddepression detectioncompresses features from short audio segments into one speaker-level decision, a step calledtemporal aggregationrarely studied on its own. Most benchmarks fix a singleself-supervised encoderand a single hand-picked layer, so a reported gain may reflect the pipeline rather than the aggregation method itself. We introduce DEPOOL, a controlled benchmark that compares six aggregation architectures with six frozenspeech backboneson an English and a Mandarin depression corpus, where each configuration learns which backbone layers matter rather than fixing one by hand. Across the resulting 72-configuration grid, a third of configurations collapse into predicting a single class for every speaker, a failure tied to the backbone as much as to the method, and the architecture that is most stable in a single-seed run becomes unreliable when training repeats across seeds.Robustnessto backbone and seed, rather than average accuracy across a single pipeline, should be a first-classbenchmarkingcriterion fortemporal aggregationinclinical speech.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.02904

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.02904 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.02904 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.02904 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Probabilistic Textual Time Series Depression Detection

arXiv cs.CL

This paper proposes PTTSD, a probabilistic framework for depression severity detection from clinical interview transcripts that models uncertainty and provides temporal interpretability, achieving competitive performance on benchmark datasets.