语义知识驱动的网络上下文广告投放高效方法研究

项目名称： 语义知识驱动的网络上下文广告投放高效方法研究

项目编号： No.61202171

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 吴宗大

作者单位： 温州大学

项目金额： 22万元

中文摘要： 随着万维网的快速普及，网络上下文广告的重要性日益显著。没有上下文广告，网络的商业价值将遭受严重损失。不同于一般性的网络文档，网络文本广告篇幅较短，包含较少的关键字，并且需要动态嵌入目标网页，因而，对精度和效率均提出了更高的要求。然而，现有的上下文广告投放方法，难以在这广告投放的准确性和高效性之间取得较好的平衡。所以，设计出一种"既准且快"的上下文广告投放方法，具有重要的科学意义和重要的应用前景。 基于语义知识库，本课题将深入研究： 1）如何利用知识库丰富的语义知识来捕获和扩充文本广告的内容特征向量，克服传统方法容易引发的语义混淆、关键词重叠率低、内容失配等问题，提高上下文广告投放的准确性。2）如何通过设计高效的语义空间映射算法，快速地获取目标网页的内容特征向量，确保上下文广告投放的高效性。 3）以及，如何将文本广告嵌入到目标网页中与广告内容最相关的位置，确保上下文广告投放的局部准确性。

中文关键词： 网络内容广告；语义知识；特征映射；内容相关度；特征空间

英文摘要： With the popularization of the World Wide Web, contextual advertising is becoming increasingly important. Without contextual ads, the Web will lose the most of its market values. Different with generic web documents, contextual ads are generally with shorter length, i.e., which contain less keyword, and have to be embedded into a targeted web page dynamically, leading to the higher requirements for precision and efficiency. However, it is difficult for existing contextual advertising approaches to balance the accuracy and efficiency in advertising. It therefore has both significant theoretic values and wide potential applications to propose an effective contextual advertising approach that is not only with good precision but also with satisfactory efficiency. In this research project, based on a semantic knowledge base, we mainly discuss the following three key issues: 1) how to leverage rich semantic knowledge to capture and enrich the feature representations of textual ads, to overcome the problems such as semantic ambiguity, low intersection of keywords, and context mismatch, consequently, improving the precision of contextual advertising. 2) how to design an effective space mapping algorithm to obtain the feature representations of web pages efficiently, consequently, ensuring the good efficiency of conte

英文关键词： contextual advertising；semantic knowledge；feature mapping；feature space；content relevance