Assessing Google Translate ASR for feedback on L2 pronunciation errors in unpredictable sentence contexts
Bibliographic record
Abstract
Following previous research into predictable sentence contexts, this study assesses the pronunciation feedback provided by Google Translate’s (GT) Automatic Speech Recognition (ASR) in unpredictable contexts. We examined the accuracy of GT transcriptions for target items recorded by male and female Quebec Francophones (QFs). The items occurred in neutral carrier sentences such that no contextual cues help ASR identify the targets. Th-initial vs t-initial (thank-tank) and h-initial vs vowel-initial (heat-eat) items were used to investigate the potential for feedback on the QF errors of th-substitution, h-deletion, and h-epenthesis, comparing real-word (thank→tank) vs nonword output (thief→tief). As with predictable contexts in our previous research, we observed high transcription accuracy for real words only. Without contextual cues, accuracy rates were lower than in predictable contexts for correctly pronounced items but higher than for incorrect pronunciations constituting real words. Unpredictable contexts are thus inferior at confirming correct pronunciation (confirmative feedback) but superior at flagging real-word errors (corrective feedback). Contrary to the anticipated ASR gender bias, female recordings showed higher transcription accuracy than male recordings. Our findings both confirm the usefulness of GT’s ASR for generating pronunciation feedback and highlight the importance of context (predictable vs unpredictable) and lexical status (real vs nonword).
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 | 0.003 | 0.017 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.004 | 0.003 |
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".