Scaling Conversational Hungarian ASR: The BEA-Dialogue+ Corpus

2026-05-29

AI Summary

This work addresses the challenge of limited publicly available conversational speech data for Hungarian automatic speech recognition (ASR), where conventional speaker separation strategies yield only 85 hours of usable training material. To overcome this bottleneck, the authors propose a novel data partitioning strategy that relaxes isolation constraints between experimenters and dialogue partners while strictly preserving primary speaker independence, thereby expanding the BEA-Dialogue+ corpus to 200 hours. Leveraging Whisper and FastConformer architectures with Serialized Output Training (SOT) fine-tuning, experiments reveal that larger-scale training without fine-tuning introduces greater difficulty, whereas SOT fine-tuning substantially reduces word error rate (WER), character error rate (CER), and context-aware metrics (cpWER, cpCER), enabling principled investigation of the trade-off between data scale and speaker overlap.

Abstract

Conversational automatic speech recognition in Hungarian is constrained by the limited amount of publicly available dialogue-style training data. The BEA-Dialogue corpus addresses this need, but its strictly speaker-disjoint train/dev/eval split reduces the usable material to only 85 hours. In this paper, we introduce BEA-Dialogue+, an expanded version of the corpus that relaxes the split criterion for experimenters and dialogue partners while preserving complete separation of the primary speakers. This results in 200 hours of transcribed natural conversations and enables a controlled study of the trade-off between additional training data and speaker overlap across the splits. We evaluate several Whisper- and FastConformer-based models on both corpus versions, including Serialized Output Training (SOT)-based fine-tuning for dialogue transcription. Our results show that the larger corpus is more challenging for models without fine-tuning, whereas SOT-based adaptation yields consistent improvements in WER, CER, cpWER, and cpCER. Overall, BEA-Dialogue+ provides a substantially larger yet still demanding benchmark for Hungarian dialogue ASR, and a practical resource for training and evaluating dialogue transcription systems.

Problem

• conversational ASR
• Hungarian
• data scarcity
• speaker overlap
• dialogue transcription

Innovation

• BEA-Dialogue+
• speaker overlap
• Serialized Output Training
• conversational ASR
• data scaling