Abstract: Live streaming e-commerce is an emerging and popular mode of online retail, where products are introduced and promoted through live streaming. Based on the signal theory and the non-verbal cue theory, this paper utilizes audio mining techniques to analyze the live streaming and related sales data from Taobao Live and investigates the influence of streamer’s voice characteristics (specifically pitch, loudness, and speech rate) on sales performance. The results show that pitch, loudness, and speech speed have a significant and robust inverted U-shaped effect on live streaming sales. Additionally, the rhythm of background music positively moderates this relationship between speech rate and sales performance. The matching of background music tempo and speech rate for a specific product category enhances sales performance. This paper further explores different influences of voice characteristics during the two-stage live streaming. The results show that the inverted U-shaped effect of speech rate is insignificant in the first stage of product explanation but becomes significant in the subsequent stage focused on convincing purchase. Overall, this paper provides a new direction for improving the sales performance of live streaming e-commerce. The findings may help live streaming e-commerce merchants and streamers to improve their online performance through strategic use of voice and background music.

Key words: live streaming e-commerce, streamer, voice characteristics, signal theory, audio mining

摘要： 基于信号理论和非语言线索理论，应用音频挖掘技术，分析了淘宝直播平台上电商主播的直播视频和带货数据，研究了电商主播的声音特征（音调、响度、语速）对其销售业绩的影响作用。研究结果表明：音调、响度、语速对直播带货销售额有显著且稳健的倒U型影响，背景音乐节奏正向调节了主播语速与带货销售额之间的倒U型关系，并且背景音乐节奏、主播语速和产品类型匹配时更有利于促进带货业绩提升。同时，进一步研究了主播声音特征在直播带货两阶段中的不同影响，发现主播语速特征在商品讲解阶段的倒U型效应不显著，而在说服购买阶段的倒U型效应显著。

关键词: 直播电商, 主播, 声音特征, 信号理论, 音频挖掘