Date created: | Last Updated:

Category: Project

Description: This study evaluates the performance of WhisperX, an open-source ASR system, and IBM Watson, a leading commercial system, on non-native English accents using the EdAcc corpus. The corpus comprises conversational audio files between speakers who share the same native language, providing a rich dataset for analyzing the impact of accent variability (Sanabria et al. 2023). The findings reveal that while WhisperX outperformed IBM Watson, the WERs for both systems remain significantly high. These results highlight the limitations of current ASR systems in handling non-native accents and emphasize the need for more diverse training datasets and system improvements to ensure inclusivity.