Recent advancements in automatic disordered speech recognition: A survey paper
Bibliographic record
Abstract
Automatic Speech Recognition (ASR) technology has recently witnessed a paradigm shift with respect to performance accuracy. Nevertheless, impaired speech remains a significant challenge, evidenced by the inadequate accuracy of existing ASR solutions. This lacking is reported in various research reports. While this lacking has motivated new directions in Automatic Disordered Speech Recognition (ADSR), the gap between ASR performance accuracy and that of ADSR remains significant. In this paper, we report a consolidated account of research work conducted to date to address this gap, highlighting the root causes of such performance discrepancy and discussing prominent research directions in this area. The paper raises some fundamental issues and challenges that ADSR research faces today. Firstly, we discuss the adequacy of impaired speech representation in existing datasets, in terms of the diversity of speech impairments, speech continuity, speech style, vocabulary, age group, and the environments of the data collection process. We argue that disordered speech is poorly represented in the existing datasets; thus, it is expected that several fundamental components needed for training ADSR models are absent. Most of the open-access databases of impaired speech focus on adult dysarthric speakers, ignoring a wide spectrum of speech disorders and age groups. Furthermore, the paper reviews prominent research directions adopted by the ADSR research community in its effort to advance speech recognition technology for impaired speakers. We categorize this research effort into directions such as personalized models, model adaptation, data augmentation, and multi-modal learning. Although these research directions have advanced the performance of ADSR models, we believe there is still potential for further advancement since current efforts, in essence, make the false assumption that there is a limited distribution shift between the source and target data. Finally, we stress the need to investigate performance measures other than Word Error Rate (WER)- measures that can reliably encode the contribution of erroneous output tokens in the final uttered message.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.006 | 0.007 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.006 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".