Open Issues Need Help
View All on GitHubKoel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
AI Summary: The task involves improving the pre-training of Grapheme-to-Phoneme (G2P) models used in a speech-to-IPA transcription system. This requires researching and potentially training G2P models that better handle accents and dialects, potentially incorporating articulatory features. The goal is to improve the accuracy of IPA transcriptions, especially for nuanced pronunciations, and explore methods to adapt the model to new languages and dialects using articulatory features.
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
AI Summary: Integrate the Boundary Loss (BL) metric into the Koel Labs' Speech2IPA model evaluation pipeline. This involves adding the BL calculation to the existing `metrics.py` script, referencing the implementation provided by Ye et al., and incorporating it into the model evaluation workflow.
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code
AI Summary: Implement Articulatory Reconstruction Loss (ARL) as a new evaluation metric for Speech2IPA models in the Koel Labs ML repository. This involves integrating the ARL calculation into the existing `metrics.py` script, potentially requiring understanding of the AAI model and the ARL calculation as described in the provided research paper.
Koel Labs innovates real-time pronunciation feedback for language learners! This repo contains the ML training, evaluation, and data processing code