基于社会化标注的搜索引擎优化研究

【作者】 张晖；

【导师】 沈洁；

【作者基本信息】 扬州大学 ， 计算机应用技术， 2008， 硕士

【摘要】 社会化标签作为web2.0中的一种重要技术,最显著的贡献在于完成了用户从单纯的信息接收者到主动的信息发布者的角色转换,深刻地影响着互联网上的信息传播模式。通过标签,其他网站或网民可以便捷地对信息进行分类。标签的作用类似于搜索技术中的关键字,是对信息的个性化描述。由于一般情况下无法对网络上海量的信息进行全文检索(只有功能强大的大型搜索引擎才能勉强做到),因而,具有标签的信息要比没有标签的信息更容易传播。网民通过“贴”标签来描述信息,创造易于被他人检索的信息,同时也通过标签的描述来查找自己感兴趣的信息。并且由于聚合内容技术与标签技术的有机结合,标准化的信息接口降低了信息传播和信息再次加工的成本,从而降低了人们在获取个性化信息上的时间成本和经济成本。通过社会化书签,可以用多个标签作为关键字,对所喜爱的网站或网页(由网络中的书签所指向)进行标注,并与其他网民分享。社会化标签的作用类似于分类式的搜索引擎,其区别在于社会化标签的创建者是一个个普通网民,而不是什么实力雄厚的公司,因而具有更强的实效性,甚至在分类上更为准确。随着web2.0技术影响力的不断扩大,web上可以挖掘的资源更加丰富。结合社会化标签来优化搜索引擎的检索效率不但能够快速准确的定位用户所需的信息,而且能够满足用户对信息个性化方面的需求,从而在很大程度上改善了用户的搜索体验,因此对这部分内容的研究也将具有重要的商业价值。论文的主要工作包括以下三个方面:(1)基于社会化标注内容的研究本文主要根据标签的语义来判断社会化标注的内容。通过比较标签与网页内容之间、标签与标签之间的语义关系,确定标签与网页内容之间、标签与标签之间的语义相似度,并根据该相似度对搜索引擎用户进行推荐,提高其对搜索的满意度,从而达到优化搜索引擎的目的。(2)基于社会化标注重要性的研究衡量标签的重要性有多个方面,本文主要考察的角度是时间。通过对社会化标签网下新的数据源“标签”的时间因素加以分析和利用,从时间的角度衡量标签的新颖程度和重要程度,提出了一种新的基于社会化标签的网页排名算法TagRank,该算法通过对网页上用户的标注行为进行挖掘,计算标签的“热度”,从而更客观的反应出标签的真实质量,以此提高网页排名的准确性。实验证明该算法是切实有效的。(3)基于社会化标注个性化的研究基于社会化标注个性化研究不需要用户的主动参与,而是通过对共现tag的数据加以过滤,对用户的标注数据聚类并获取用户的偏好,达到对tag进行层次上分类的效果,从而得到一个关于用户的特征,根据该特征向用户有的放矢的推荐,最终成功的为用户提供个性化服务。更多还原

【Abstract】 As a very important technique of web2.0, the most prominent contribution of social annotations is that it successfully makes web users turn from being simplex information acceptors to active information promulgators and profoundly influence the information transmitting mode on the web. With tags, other websites or users can conveniently classify the information. The function of tags is similar to which of the keywords in search technique, it’s the personalized description to information. In the usual situation, it’s very hard to carry through whole-length text retrieval (only those powerful search engines can constrainedly achieve), so the annotated information are more easily transmit than the un-annotated. Web users describe the information and create easy retrieval by the action of annotating, and they can look for the information which they are interested in by the description of tags. Because of the organic combination of polymerization content technique and annotation technique, standardized information interface reduce the cost of information transmitting and re-process, consequently, reduce the cost of time and economy when people want to acquire personalized information. With social bookmarks, we can use multi-tags as keywords to annotate the websites and webpage(be point to by bookmarks on the web ) which we are fond of, and share them with other web users. The function of social annotations is similar to classifying search engines, the difference between them is that the creators of social annotations are normal web users rather than powerful corporations. So it has more actual effect and classify more accurately.With the influence extension of web 2.0 technique, the resource which can be digged on web is more abundance. Utilizing social annotations to optimize the retrieval efficiency of search engines can not only quickly and accurately go to the information which the users need but also satisfy the users’need of personalized information. Thereby, it obviously improves the search experience of the users. So the research on this aspect has significant commercial value.The major work in this paper include the following three parts:(1) the research based on the content of social annotationIn this paper, we judge the content of social annotations mainly by the semantic of tags. By comparing the relation between tags、tags and webpage, we measure the semantic similarity of them. And commend to the users according to this semantic similarity, consequently, achieve the aim to optimize search engines.(2) the research based on the significance of social annotationThere are many aspects to scale the significance of social annotation. In this paper, we mainly consider the time factor. We analyze and utilize the time factor of the new data source“tag”to weigh the importance and novelty of tags and propose a new algorithm named TagRank based on social annotations for page ranking. This algorithm digs the annotation behavior of the web users, calculates the“heat”of the tags, consequently, it can response the true quality of tags more externally and improve the veracity of page ranking. The experiment shows that our algorithm works effectively.(3) the research based on the personalization of social annotationThe research based on the personalization of social annotation doesn’t need the active participation of users, but by the filtration of co-tag data, we can cluster the users’annotation data and acquire their preference to classify tags in level, consequently, get a characteristic about the user, and purposefully commend to the user according to this characteristic, finally provide personalized service to the user successfully.更多还原