管理学系学术讲座

讲座题目：From Voice to Purchase: An Emotion-Enhanced Dual-Involvement Attention Model for Early Customer Prediction

时间：2025年6月12日（周四） 13:30——17:30

地点： 宁远楼211

主讲人：王聪

 

主讲人简介：

王聪，北京大学光华管理学院管理科学与信息系统系助理教授。于清华大学经济管理学院取得管理学博士学位，于北京大学信息管理系、国家发展研究院取得管理学、经济学双学士学位，曾在Carnegie Mellon University（卡耐基梅隆大学）从事博士后研究工作。学术研究聚焦于机器学习、数据挖掘等技术方法与管理问题的交叉点上，根据不同管理问题需求及其数据特点进行相应的分析方法设计，以提供精准高效的解决方案。目前主要关注于电子商务、金融科技、智慧医疗等领域的决策支持方法设计研究。研究成果发表于国内外知名学术期刊，如 Information Systems Research, INFORMS Journal on Computing, 管理世界等。

讲座简介：

Customer purchase prediction is vital for effective resource allocation and strategic sales planning. Voice interactions, a rich source of behavioral signals reflecting customer intent, pose challenges due to their unstructured nature. To address this, we introduce the Emotion-enhanced Dual-Involvement Attention （EDIA） model for early prediction of purchase decisions from voice data. Grounded in involvement theory, EDIA captures two key dimensions: cognitive inquiries and emotional expressions. Leveraging large language and speech models, EDIA extracts multimodal features, combining linguistic and acoustic signals. It designs a dual attention-like mechanism to model high- and low-involvement patterns and integrates an emotion-aware gating module to enhance prediction accuracy. Tested on real-world automotive sales calls, EDIA outperforms state-of-the-art methods in identifying high-potential customers early. By enabling precise engagement strategies, EDIA optimizes resource allocation, improves conversion rates, and maximizes return on marketing investment, offering a robust solution for sales operations.