Content recommendations in a live customer environment

在当今信息超载的时代，信息生产者和信息消费者都面临着重大挑战。对于信息生产者来说，如何使自己的信息从大量的信息中脱颖而出，吸引更多用户的注意，是一项艰巨的任务。对于信息消费者来说，如何从海量信息中快速准确地找到感兴趣的信息也是一个挑战。推荐制度在一定程度上试图解决这些矛盾。它可以连接用户和信息。一方面，它帮助用户找到有价值的信息。另一方面，它使信息呈现给那些对它感兴趣的用户。从而实现信息提供者和用户之间的双赢。本项目拟开发几种相互竞争的内容推荐算法，结合个人用户的兴趣和参与度、受欢迎程度、新鲜度、内容类型等数据，然后在现场客户网站上测试其有效性。创新在于三个方面。首先，本研究将提出一种有效的方法来处理由基于关键字的向量空间转换而来的高维稀疏数据。其次，本课题将提出一种考虑关键词特征之间相关性的特征排序方法。这样既可以提取主效应强的特征，又可以捕获主效应相对较小的特征

In the current era of information overload, both information producers and consumers are all facing a significant challenge. For information producers, how to make their own information stand out from the substantial information and attract the attention from more users is a difficult task. For information consumers, how to quickly and accurately find the interested information from the massive information is also a challenge. Recommended system attempts to solve these contradictions to some extent. It can connect users and information. On the one hand, it helps users find the valuable information. On the other hand, it makes the information presented to those users who are interested in it. Thus, a win-win situation can be achieved between the information providers and users. This project proposes to develop several competing content recommendation algorithms, incorporating individual user interests and engagement, popularity, freshness, content type and other data, and then test their effectiveness on live customer websites. The innovation lies in three perspectives. Firstly, the proposed research will propose an efficient approach to deal with the high-dimensional sparse data converted from the keyword-based vector space. Secondly, this project will propose a feature ranking method taking into account the correlation among keyword features. Thus, it can not only extract the features with strong main effect, but also capture the features with a relatively small main