Methods Design Of Search Engine Web Page Relevance Assessment And Application In Rank Model

With the rapid development of internet, it is difficult for ordinary internet users to find the required information from the internet. In response to this demand, the search service came into being. According to the web's quality and the site's authority and other factors, search engine technology will rank web pages comprehensively. When users retrieve the needed information, the relevant results have priority to show, which has a major significance for ordinary Internet users. Search engine has the goal of meeting the users' information needs mostly, but in the actual retrieval, for some queries with low frequency, or less resource online, it was difficult for search engine to retrieve the good result and return to the users, which brought the users bad experience.Relevance assessment is a basis and core of the steps which enhance search quality, the assessment can find the defects of the current technology, and then improve search engine quality. Search engine relevance assessment results will affect hundreds of millions of users'search experience. Through relevance assessment of unpopular query, analyzed the current problems in rank model, thereby improved the user experience on the unpopular query. My main work in this article is as follows:1. Designed the relevance 5-rank assessment method based on the relevance 4-rank assessment method, which is the important innovation of this article;2. Through Delphi experiment, get the user's relevance assessment data, then made the relevance 4-rank and 5-rank assessment data fitting linear curve, then proved that the relevance 5-rank assessment method was more accuracy for the unpopular query assessment;3.Organized users to assess 200 unpopular queries with relevance 5-rank assessment method, including Baidu 1000url, Google 1000url and ideal data 1000url, and utilized the marked results to quantify the gap between Baidu and Google, and the gap between Baidu and the ideal situation. Analyzed the specific cases, and gave the improvement direction of the rank model.The relevance 5-rank assessment method provided more accurate data supporting for rank model; thereby improved the users'retrieval experience.