Collective user behavior analysis for social intelligence

社交智能的集体用户行为分析

• 批准号: 531376-2018
• 负责人: Du, Weichang
• 金额: $ 1.82万
• 依托单位: University of New Brunswick
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=649630
• 关键词: Collective; user; behavior; analysis; social

The focus of this research and development project is on life event mentions on social media. Our work will**focus primarily on aspects that relate to life event prediction. While a large body of research has focused on**multimedia content such as images and videos to automatically retell a series of life events, little work has been**done on streaming textual content in online social networks for the purposes of life event prediction. This task**is challenging due to the short, informal and noisy characteristics of social media posts such as tweets. Further**and more importantly, the task also suffers from the "sparsity problem," as social media is dominated by**information other than life events. Research in this area has effectively used various types of features including**social interaction, syntactic, semantic, and neural embeddings to train classifiers with reasonable performance.**The objective of this project will be to work in close collaboration with ThinkCX, the industrial partner, and to**develop a multidimensional framework for exploring how life events can be predicted based on streaming**social textual content. This will enable the industrial partner to build tools that gain deeper insight into the**users' preferences and interests based on their life events, resulting in higher customer satisfaction and**customer engagement. The outcomes of this project will enable the industry partner to fine-tune its**understanding of the end-users and hence potentially increase its customer base, leading to growth in revenue**and the hiring of more employees in Canada.

该研究和开发项目的重点是社交媒体上的生活事件提及。我们的工作将主要集中在与生活事件预测有关的方面。虽然大量的研究都集中在多媒体内容，如图像和视频自动复述一系列的生活事件，很少有工作已经 ** 做在线社交网络中的流文本内容的生活事件预测的目的。这项任务 ** 具有挑战性，因为社交媒体帖子（如推文）具有简短、非正式和嘈杂的特点。此外，更重要的是，这项任务还存在“稀疏性问题”，因为社交媒体被生活事件以外的信息所主导。这一领域的研究已经有效地使用了各种类型的特征，包括 ** 社会交互，句法，语义和神经嵌入，以训练具有合理性能的分类器。该项目的目标是与行业合作伙伴ThinkCX密切合作，并开发一个多维框架，以探索如何根据流式社交文本内容预测生活事件。这将使工业合作伙伴能够构建工具，根据他们的生活事件更深入地了解 ** 用户的偏好和兴趣，从而提高客户满意度和客户参与度。该项目的成果将使行业合作伙伴能够微调其对最终用户的理解，从而有可能增加其客户群，从而增加收入 ** 并在加拿大雇用更多员工。