RadioNews-BBC

This dataset is released as part of the paper "Exploring Pre-Trained Neural Representations for Audio Topic Segmentation" and it includes embeddings extracted from non-overlapping 1-second audio portions from various news shows from BBC radio channels. Each audio file has been anonymised by labelling it with a randomised label. We release 7 type of embeddings coming from different pre-trained architectures and, where applicable, for 3 of these embedding type we further release 7 sub-folders containing the actual audio embeddings for each file. These sub-folders contain the embeddings obtained with different pooling strategies described in the original paper: as openL3, Wav2Vec2 and CREPE are trained to output multiple embeddings for each 1-second frame the pooling strategies reduce those multiple embeddings to one per frame. The pooling strategy can have a huge impact on final model performance.<br> Finally, we release the ground truth for each audio file as a pickle file containing a python dictionary, where the keys are the same identifiers used to name the embeddings (without the .npy extension). The ground truth were produced by manual annotators and they represent whether each 1-second frame is a topic boundary (i.e. a topic shift happens in or at the end of the frame) or not, where 1 corresponds to topic boundary and 0 to in-topic frames (i.e. non-boundary). Below we describe in more details the structure of our dataset: - PodcastUniform1: Parent directory containing all the other directories and files. Uniform 1 indicates the initial segmentation methodology, consisting of non-overlapping 1-second chunks of audio. The parent directory includes the following subdirectories: - - openL3: a folder of folders, one for each pooling strategy, each containing numpy arrays, one for each audio source file, including the relative openl3 embeddings. - - x-vectors: a folder of numpy arrays, one for each audio source file, including the relative x-vector embeddings. - - ecapa: a folder of numpy arrays, one for each audio source file, including the relative x-vector embeddings. - - Wav2Vec2: a folder of folders, one for each pooling strategy, each containing numpy arrays, one for each audio source file, including the relative wav2vec2 embeddings. - - CREPE: a folder of folders, one for each pooling strategy, each containing numpy arrays, one for each audio source file, including the relative CREPE embeddings. - - prosodic: a folder of numpy arrays, one for each audio source file, including the relative prosodic embeddings. - - mfcc: a folder of numpy arrays, one for each audio source file, including the relative MFCC embeddings. - - labs_dict.pkl: a pickled file (Protocol version 5) containing the topic segmentation ground truth. It consists of a dictionary where each key is the identifier assigned to the original audio file- and the value associated is a list of 0 and 1s of length equal to the corresponding embedding containing the same name. The elements in each list indicate whether the corresponding embedding constitutes a topic boundary (1) or not (0) and it is therefore used to train and test a topic segmentation model. For example, in the Non-news dataset the key "24260" contains the ground truth for all the numpy array files containing the identifier "24260" in the same dataset (e.g. openl3/_mean_/24260.npy, prosodic/24260.npy, etc.). This work is licensed under the Creative Commons Attribution-NonCommercial 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/3.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.本数据集随论文《探索用于音频主题分割的预训练神经表征》一同发布，其中包含从BBC广播电台各新闻节目中提取的非重叠1秒音频片段的嵌入向量。每个音频文件均已完成匿名化处理，通过随机生成的标签进行命名。我们共发布7类源自不同预训练架构的嵌入向量；针对其中3类嵌入向量，我们额外提供了7个子文件夹，用于存储各音频文件的实际嵌入向量。这些子文件夹包含了原文所述的不同池化策略生成的嵌入向量：由于openL3、Wav2Vec2和CREPE模型会为每个1秒帧输出多组嵌入向量，池化策略可将这些多组嵌入向量整合为每帧一组的单一嵌入向量。池化策略对最终模型性能具有显著影响。 最后，我们还发布了各音频文件的主题分割真实标注（ground truth），其存储为pickle格式文件，内含Python字典，字典的键与嵌入向量文件的命名标识符一致（不含.npy扩展名）。该真实标注由人工标注人员生成，用于标注每个1秒帧是否为主题边界（即该帧内或帧末尾发生主题切换），其中1代表主题边界，0代表主题内帧（即非边界帧）。 下文将详细描述本数据集的结构： - PodcastUniform1：包含所有其他目录与文件的根目录。其中“Uniform 1”指代初始分割方法，即采用非重叠的1秒音频片段进行切分。该根目录包含以下子目录： - openL3：多层文件夹结构，每个池化策略对应一个子文件夹，每个子文件夹内包含若干numpy数组文件，每个文件对应一个音频源文件，存储对应的openL3嵌入向量。 - x-vectors：存储若干numpy数组文件的文件夹，每个文件对应一个音频源文件，存储对应的x-vector嵌入向量。 - ecapa：存储若干numpy数组文件的文件夹，每个文件对应一个音频源文件，存储对应的x-vector嵌入向量。 - Wav2Vec2：多层文件夹结构，每个池化策略对应一个子文件夹，每个子文件夹内包含若干numpy数组文件，每个文件对应一个音频源文件，存储对应的Wav2Vec2嵌入向量。 - CREPE：多层文件夹结构，每个池化策略对应一个子文件夹，每个子文件夹内包含若干numpy数组文件，每个文件对应一个音频源文件，存储对应的CREPE嵌入向量。 - prosodic：存储若干numpy数组文件的文件夹，每个文件对应一个音频源文件，存储对应的韵律嵌入向量。 - mfcc：存储若干numpy数组文件的文件夹，每个文件对应一个音频源文件，存储对应的梅尔频率倒谱系数（MFCC）嵌入向量。 - labs_dict.pkl：采用协议版本5序列化的pickle文件，存储主题分割真实标注。该文件内含一个字典，字典的每个键为原始音频文件的标识符，对应的值为与对应嵌入向量长度一致的由0和1组成的列表，列表中的每个元素代表对应的嵌入向量是否对应主题边界（1）或非主题边界（0），可用于训练和测试主题分割模型。例如，在非新闻数据集中，键“24260”对应所有包含标识符“24260”的numpy数组文件的真实标注（如openl3/_mean_/24260.npy、prosodic/24260.npy等）。 本作品采用知识共享署名-非商业性使用3.0未移植许可协议进行许可。若要查看该许可协议副本，可访问http://creativecommons.org/licenses/by-nc/3.0/，或致函至Creative Commons, PO Box 1866, Mountain View, CA 94042, USA。