Audio conversations segmented and transcribed for speech recognition

- 100 hours of Norwegian bokmål speech data of 100 people in conversations between two people. - Each recording is segmented by: speech, noise, and music. - Speech segments transcribed for speech recognition. - Speakers are identified throughout the entire dataset using a unique id. - Roughly 50% female and 50% male speakers. - Conversations cover varying general topics.

搜集汇总

背景与挑战

背景概述

该数据集包含100小时的挪威语（Bokmål）对话音频，涉及100名说话者，每段录音被分割为语音、噪声和音乐部分，语音部分已转写用于语音识别。说话者通过唯一ID标识，男女比例大致均衡，对话主题涵盖广泛的一般性话题。