Research On Presentation And Evaluation Of Search Results

Search results evaluation is the core component of search engine evaluation,which is important to evaluate user experience and iteratively improve search engine.At present,search results evaluation mainly focuses on the content of search results,but ignore the way in which search results are presentationed or assume that all search results are presentationed in the same way.As main component of search results,text search results have been generated and presentationed in the way of query oriented all along,and the presentation modes have been solidified into highlighting of title,query words and URLs.This method faces some challenges in current Internet environment:(1)number of users is increasing,search scenarios and tasks are diverse,keyword-based queries are often short or ambiguous and there are often deviations between search results presentation based on query words highlighting and users' real query needs;(2)users expect search engines to fulfill their requirements of "the answer is the question",but there are deviations between query-oriented presentation and users' expectations of knowledge acquisition;(3)there are deviations between content evaluation of search results in the same and single presentation mode and users' overall cognition of search results.Considering challenges above,this thesis mainly studies the methods of presentation and evaluation of search results.From the perspective of meeting users' information needs,the presentation effect of search results is evaluated,and the presentation mode of search results is automatically optimized.This thesis carries out the following three specific studies: search results presentation and evaluation based on users' query intention,presentation and evaluation of question query oriented to answer perception,and automatic evaluation method of search results presentation effect.In the previous two researches,methods of user survey and user experiment are applied,with which the users' preference for search results presentation and the impact of presentation mode on users' search behavior are analyzed qualitatively and quantitatively,in condition that query submitted by users are key words or questions.We use machine learning algorithm to realize search results presentation based on user intent,which can be used to optimize current presentation mode of search results,and can provide a "golden standard" for the evaluation of search results presentation effect.The above research results extend the existing research of search engine results evaluation,and deepen our understanding of the impact of search results presentation on users' search behavior.On this basis,the last research realizes automatic evaluation of the presentation effect of search results,which can greatly reduce the required human and economic costs,and shorten the feedback cycle of evaluation.