利用语音基础模型中的语言选择性维度检测语码转换
原标题：Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models
AI 摘要
Apple 研究团队提出一种轻量级语码转换（CS）检测方法，直接利用预训练语音编码器中的帧级语言信息判断一句话中是否同时出现两种语言，而无需依赖多语言转录。该方法通过单语录音识别语言选择性隐藏维度，并聚合其帧级激活进行 CS 预测，再用小规模 CS 数据集调整决策规则。实验显示 Whisper-small 编码器在 CS-FLEURS 的 14 个语言对上平均 F1 达 0.798，优于 CNN 的 0.776 和 HuBERT 线性基线的 0.751，表明预训练语音编码器已编码识别 CS 所需信息，可用于构建更具代表性的多语言语料。
正文节选
Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models AuthorsJie Chi, Zakaria Aldeneh, Masha Fedzechkina, Maureen de Seyssel, Natalie Schluter Code-switching (CS), the alternation between languages within an utterance, is common in multilingual speech but underrepresented in pretraining corpora. Increasing CS coverage requires finding relevant utterances in large speech collections so that they can be included in training data. One way to find such utterances is