Earnings Call

该数据集包含了2017年标普500公司每个季度的盈利电话会议，去重后共有169次会议记录。数据集中包含了诸如公司名称、产品名称等特定领域的详细信息。由于包含多个发言人和不稳定的录音环境等因素，这些录音的音频质量相较于专业录制的数据集有所下降。该数据集的规模为169次电话会议，分为40小时的训练集、10小时的有效集和10小时的测试集，其任务是针对特定领域的自动语音识别（ASR）。

This dataset contains quarterly earnings conference calls of S&P 500 companies in 2017, with a total of 169 deduplicated meeting recordings. It includes detailed domain-specific information such as company names and product names. The audio quality of these recordings is inferior to that of professionally recorded datasets, owing to factors including multiple speakers and unstable recording environments. Comprising 169 conference calls in total, the dataset is split into a 40-hour training set, a 10-hour validation set, and a 10-hour test set, tailored for the task of domain-specific automatic speech recognition (ASR).

搜集汇总

背景与挑战

背景概述

该数据集包含2017年S&P 500公司的收益电话会议文本和音频记录，用于研究CEO的言语和声音特征对股票波动性的预测作用。数据集提供了详细的处理方法和使用说明，适合金融和自然语言处理领域的研究。