Fleurs Openslr42 Mpwt: Khmer Speech Recognition Corpus

KrorngAI combined and cleaned three speech datasets to create this Khmer-language audio resource. The creator manually corrected transcriptions for over 2,000 rows and normalized text by removing invisible whitespace and processing special characters. This dataset was last updated on October 30, 2025.

Use Cases

• Train automatic speech recognition models based on the cleaned and normalized Khmer transcriptions.
• Benchmark multilingual speech systems based on the combined data from Google FLEURS and OpenSLR42.
• Develop text normalization tools for Khmer based on the described processing of special characters, numbers, and dates.

Strengths

• Includes over 2,000 manually corrected transcriptions, suggesting a focus on data quality.
• Integrates data from three established sources: Google FLEURS, OpenSLR42, and a cleaned Khmer speech corpus.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic or source bias inherent to the original datasets.

Provenance

Source

Combines google/fleurs, openslr/openslr42, and cleaned seanghay/khmer_mpwt_speech.

Collection Method

Manual correction of transcriptions and automated text normalization.

Freshness

Last updated 2025-10-30 04:54:33; freshness should be verified.