Linguistic challenges in automatic speech recognition (ASR)

July 7, 2026Open Access

View Full Paper

Ask AI

Bookmark

Share

Key Points

The study investigates the performance of automatic speech recognition systems in transcribing non-native English speech in three modes.
Assessment of ASR transcripts compared to verbatim transcriptions of the same source speech.
Utilization of a sequential explanatory design.
Analysis of error categories using the normalized error rate (NER) to quantify transcription errors.
Reading a scripted speech achieved the highest agreement with ASR transcripts (NER 65.22).
The non-interpreted monologue showed a higher error rate (NER 164.47).
Consecutive interpretation resulted in the highest error rate due to adaptive interference (NER 181.10).

Examines ASR transcription performance across different non-native English speech types, indicating significant discrepancies.

No abstract yet. The journal hasn't published one we can show.

No takes yet. Share an insight, caveat, or question.

Sandra Anita Dahl (2026) studied this question.

View Full Paper

Ask AI

Bookmark

Share