Web Retrieval Based On Social Annotation

With the rapid development of Web2.0technology in recent years, the new technologyand new application greatly promote the rapid development of the Internet. The Web2.0technology makes the Internet with the social characteristics. Users are not only the recipientof web content, but also the participant. The web content in Web2.0has the reusedcharacteristics, the network user can forward, share and collect web content. How to optimizeweb retrieval by using social characteristics of Internet has attracted lots of scholars. Socialannotations that rose accompanied with Web2.0have been an aspect of researching socialsearch. The technology greatly enhances users’ enthusiasm in participating in the Internet.Social annotations support the organization storage, management and online Bookmarksfor users. There are many Social annotation services, such as Delicious for Bookmarks, Flickrfor sharing pictures and YouTube for sharing videos. Based on researching a lot of Socialannotations, I find they have tow characteristics: one, social annotation are great summaries ofwebpage. Users describe the Web pages in the metadata form, so that they can understand thecontent of the page without reading it. Another one, Social annotations can compute similaritybetween social annotations. Many social annotations may be marked to the same webpage,and it is deduced that they have semantic relevance between them depending on webpage.Based on the tow characteristics, I propose an algorithm named AnnoSim that computeswebpage relevance. The following introduces my work in detail:To start with, get social annotations. The paper uses social annotations of Delicious toresearch. Delicious is a social annotation site, It stores social annotations with a plainstructure. Users can share web pages to Delicious. When users share web pages to Delicious,it can get social annotations by users adding keywords in metadata form. Social annotationswhich are in the same URL will be converged together. The paper extracts social annotationswith HtmlParser and stores them with Mysql.Next, I propose an algorithm called AnnoMatchSim that calculates the relevance of thewebpage with annotations. The algorithm is similar to SVM that uses the content of page tocalculate the relevance of document, and it calculates the relevance of webpage with socialannotations. The algorithm named SVM calculates the weight of the Term in the document,and then calculates the cosine between the queries and documents. AnnoMatchSim is similarto SVM and it calculates the weight of the annotation in the document, and then calculates thecosine of the queries and documents. The value of cosine is the relevance of the document.In addition, I propose an algorithm called AnnoS that calculates the similarity betweenthe annotations. Social annotations have semantic relevance between them depending onwebpage. And the more the tow annotations mark the same webpage, the more relevant they are; otherwise, the less relevant they are.Last but not least, I propose an algorithm called AnnoSim that integrates AnnoMatchSimand AnnoS to calculate the webpage relevance. It is not satisfied with AnnoMatchSim in thecondition of romantic relevance. To improve AnnoMatchSim, I propose an algorithm calledAnnoSim that integrates AnnoMatchSim and AnnoS.To research social annotations, the paper selects50words, and the number of URL is20000.The paper checks the result of optimizing web search by using social annotations to.The result shows that: social annotations that mark the same webpage are similar,AnnoMatchSim can optimize web search, AnnoSim can optimize web search more thanAnnoMatchSim.