基于图排序的社会媒体用户的消费意图检测  

Detecting consumption intention based on graph ranking in social media

作　　者：刘挺[1] 付博[1] 陈毅恒[1] 

机构地区：[1]哈尔滨工业大学社会计算与信息检索研究中心,哈尔滨150001

出　　处：《中国科学：信息科学》2015年第12期1523-1535,共13页Scientia Sinica(Informationis)

基　　金：国家自然科学基金面上项目(批准号:61472107)、国家自然科学基金青年项目(批准号:61202277)、国家自然科学基金重点项目(批准号:61133012)资助;国家重点基础研究发展计划(973计划)(批准号:2014CB340503)

摘　　要：消费意图是指用户为满足某种需要,在一定购买动机的支配下,通过文本内容表达出对产品或服务的购买意愿.消费意图检测旨在推断用户在文本中是否表达出消费意图的含义.我们定义微博中的消费意图必须包含两个重要元素,分别是消费意图触发词和消费意图对象(即需要购买的产品),这两种元素直接引发用户的购买意愿,是决定用户消费意图的重要特征.本文提出了基于弱监督的图排序算法,该方法适用于数据总量较大、已标注数据量相对较小的情形中,并且可以使未标注数据和标注数据同时参与到图排序算法的学习过程中.实验结果表明,所采用的图排序方法对于消费意图检测是行之有效的.Consumption intention in a microblog indicates either an imminent or a future purchase, and this indicates a buying intention. Consumption-intention detection aims to determine the willingness of a user to make a purchase. The consumption intention in a microblog should contain two important elements, namely the trigger words and the consumption intention target（the product to be bought）. These two elements directly indicate a user＇s purchase intention, which is an important feature of the user＇s consumption intention. In this paper, we propose to study the problem of detecting consumption intention in microblogs based on graph ranking. This method can be applied to cases where there is a large amount of data, the amount of labeled data is relatively small, and all data can be involved in the learning process of the graph-ranking algorithm. Experimental results show that the graph-ranking-based method is effective for detecting consumption intention.

关 键 词：消费意图 消费意图检测 社会媒体 图排序 弱监督 

分 类 号：F713.55[经济管理—市场营销]