搜索引擎用户查询的广告点击意图分析

Ads-clicking intent analysis of search engine queries

摘要:

搜索引擎广告点击率的多少直接影响搜索引擎的收入,而深入分析用户查询的广告点击意图则是提高广告点击率的基础性工作．针对与此,基于商用搜索引擎的用户查询点击日志,统计分析了搜索引擎用户查询的广告点击率,提出基于查询词内容匹配和基于贝叶斯分类的两种方法预测搜索引擎用户查询的广告点击意图．在大规模的真实用户查询点击日志上的实验结果表明,所提出的方法能够预测查询的广告点击意图,将广告投放的精度从3.0％提高到36.8％,广告投放的平均F-measure值从0.060提升到0.408．通过广告点击意图预测,有效缩小了广告投放范围,并适用于在线广告意图的实时预测．

Abstract:

Click through rate (CTR) on sponsored search ads determines the search engine’s revenue, thus analysis on users’ ads-clicking intent is one of the fundamental work to improve CTR. Based on the search logs provided by a Chinese search engine, this paper presents statistical analysis of ads clicks, and further proposes two methods to predict ads-clicking intent of query, namely query content match based prediction and Bayesian classification, respectively. Experimental results on large scale real data show the improvements from 3.0％ to 36.8％ in precision and from 0.060 to 0.408 in F-measure on sponsored search ads delivery. The proposed methods are capable of predicting the intent of user queries and enhancing the effect of search engine advertising, and are also applicable for online prediction of advertising click intent of user queries.