espnet/kosp2e-asr-ko

KoSP2E ASR Recipe

This is the ESPnet2 recipe for the KoSP2E (Korean Speech Perception and Production Experiment) dataset.

---

Overview

The KoSP2E dataset is a large-scale Korean speech corpus designed for speech perception and production experiments. This recipe provides a full ASR pipeline using ESPnet2 with both Transformer and Conformer architectures.

---

Results

Environment

• Date: Mon Nov 10 20:35:20 UTC 2025
• Python: 3.10.19
• ESPnet: 202509
• PyTorch: 2.9.0+cu128
• Model: Conformer (BPE=2000)
• Decode: Transformer LM (valid.acc.ave)

WER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test	2320	22337	77.1	20.4	2.6	4.4	27.4	76.4

CER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test	2320	84267	92.5	5.7	1.8	1.7	9.2	76.4

TER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test	2320	65361	89.4	8.6	2.0	2.1	12.7	76.4

---

References

• KoSP2E paper: https://arxiv.org/abs/2107.02875

---