BanglaTurn：孟加拉语语音中话轮结束检测的基准与基于Whisper的模型
BanglaTurn: A Benchmark and Whisper-Based Model for End-of-Turn Detection in Bangla Speech
- Vivasoft Limited（维瓦索夫特有限公司）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本文提出BanglaTurn语料库及基于Whisper的模型，用于孟加拉语对话语音的话轮结束检测，在测试集上准确率达84.33%，显著优于基线，并降低了假阴性率。
AI中文摘要：
本文介绍了BanglaTurn，一个用于孟加拉语对话语音中话轮结束检测的语料库，以及在该语料库上训练的模型。该语料库包含35,374个样本，每个样本为3至15秒的播客语音，通过结合说话人日志和LLM处理来标注话轮状态，每个标签随后由人工标注员进行检查。该模型将Whisper编码器与任务特定的分类头配对。在一个从保留播客中抽取的类别平衡测试集上，该模型达到了84.33%的准确率（95%置信区间80.3至88.1），而Smart-Turn v3基线的准确率为69.28%，并将假阴性率从51.57%降至7.55%，但代价是假阳性率更高。我们报告了编码器层微调、多尺度池化和INT8量化各自带来的贡献，且在CPU上的端到端延迟保持在165至191毫秒之间。
英文摘要：
This paper presents BanglaTurn, a corpus for end-of-turn detection in Bangla conversational speech, and a model trained on it. The corpus holds 35,374 samples of 3 to 15 s of podcast speech, labelled for turn state by combining speaker diarization with an LLM pass, with every label then checked by a human annotator. The model pairs a Whisper encoder with task-specific classification heads. On a class-balanced test set drawn from a held-out podcast, it reaches 84.33% accuracy (95% CI 80.3 to 88.1) against 69.28% for the Smart-Turn v3 baseline, and lowers the false negative rate from 51.57% to 7.55% at the cost of a higher false positive rate. We report what encoder layer fine-tuning, multi-scale pooling and INT8 quantization each contribute, and latency stays within 165 to 191 ms end to end on CPU.
