Guidelines are required for accurate and consistent transcription of speech corpora, especially when they contain more challenging, e.g. spontaneous or under-resourced speech. This paper presents a workflow and guidelines for transcribing spontaneous and under-resourced child speech in AusKidTalk, the first Australian English child corpus. Speech samples were elicited using a story-telling task and are 3.5 minutes long per child on average. Orthographic transcriptions were generated using automatic speech recognition (ASR) tools and corrected manually. A novel hand-correction protocol consisting of guidelines, hand-correction interface, and ground truth transcriptions together with consistency metrics were developed. Nine annotators submitted hand-corrections for 261 children’s story-telling task, and 25 ground truth tasks. Manual correction was 11-fold of speech time with a 3.5-minute-long story-telling task corrected in approximately 40 minutes. Efficiency is attributed to the quality of automatic transcription with 23% word error rate. Manual correction was accurate with annotators achieving consistent results on 15/25 ground truth submissions. Most inconsistent ground truth submissions were caused by a single, challenging ground truth task. These results show that our workflow yields efficient and accurate transcriptions, although transcriptions of potentially more challenging narrative tasks (e.g., elicited from younger children) might require further corrections.

Large Language Models (LLMs) have been increasingly adopted for health-related tasks, yet their performance in depression detection remains limited when relying solely on text input. While Retrieval-Augmented Generation (RAG) typically enhances LLM capabilities, our experiments indicate that traditional text-based RAG systems struggle to significantly improve depression detection accuracy. This challenge stems partly from the rich depression-relevant information encoded in acoustic speech patterns — information that current text-only approaches fail to capture effectively. To address this limitation, we conduct a systematic analysis of temporal speech patterns, comparing healthy individuals with those experiencing depression. Based on our findings, we introduce Speech Timing-based Retrieval-Augmented Generation, SpeechT-RAG, a novel system that leverages speech timing features for both accurate depression detection and reliable confidence estimation. This integrated approach not only outperforms traditional text-based RAG systems in detection accuracy but also enhances uncertainty quantification through a confidence scoring mechanism that naturally extends from the same temporal features. Our unified framework achieves comparable results to fine-tuned LLMs without additional training while simultaneously addressing the fundamental requirements for both accuracy and trustworthiness in mental health assessment

Depression is a critical concern in global mental health, prompting extensive research into AI-based detection methods. Among various AI technologies, Large Language Models (LLMs) stand out for their versatility in healthcare applications. However, the application of LLMs in the identification and analysis of depressive states remains relatively unexplored, presenting an intriguing avenue for future research. In this paper, we present an innovative approach to employ an LLM in the realm of depression detection, integrating acoustic speech information into the LLM framework for this specific application. We investigate an efficient method for automatic depression detection by integrating speech signals into LLMs utilizing Acoustic Landmarks. This approach is not only valuable for the detection of depression but also represents a new perspective in enhancing the ability of LLMs to comprehend and process speech signals. By incorporating acoustic landmarks, which are specific to the pronunciation of spoken words, our method adds critical dimensions to text transcripts. This integration also provides insights into the unique speech patterns of individuals, revealing the potential mental states of individuals. By encoding acoustic landmarks information into LLMs, evaluations of the proposed approach on the DAIC-WOZ dataset reveal state-of-the-art results when compared with existing Audio-Text baselines.