Metamorphosis Test: Research On Search Engine Performance Based On User Retrieval Habits

Search Engine refers to the system that retrieves relevant information from the Internet according to certain keywords entered by the users.After the information is organized and processed,the search engine presents it to users.Search engine is the main interface for users to find the information they need online.Due to the huge amount of information on the Internet,it is impossible to verify whether the search engine returns the correct result for the users' query.There is no objectively recognized Oracle,so it is difficult to test it using traditional testing methods.Metamorphic testing is a testing technique that can be used to verify the functional correctness of software without the ideal Oracle.Therefore,this paper applies metamorphic testing to the test of search engine.Search engine is a typical information retrieval system,so the retrieval performance of search engine should be measured by precession and recall.However,in the process of search engine performance evaluation,precession and recall can't be calculated.On the one hand,search engines usually return thousands of results for the user's query,making it extremely time-consuming to calculate the amount of results related to users' queries in the search results.On the other hand,the total amount of information related to users' query on the Internet can't be known.Later,some quality evaluation criteria for web search engines were proposed by some experts,but none of them were widely used.For users,search engines' retrieval performance goes well beyond the pure classification of the result in relevant or non-relevant ones.The relevance judgements may be an important point in the evaluation of search engines,but surely not the only one.Although there have been many studies have been conducted on search engine quality and the behavior of search engine users,they are separate studies,and there are few studies focusing on an overall picture of search engine quality from the user perspective.This paper uses search logs to study users' retrieval habits,and then combines users' retrieval habits and search engine functional properties to define metamorphic relations.Based on the property of each metamorphic relation,the ability to extract web page information and understand users' intent,and search operators Supporting ability,semantic retrieval ability and semantic error correction ability are the criteria for measuring the performance of search engines.At the same time,appropriate test cases are defined based on users' retrieval habits.Finally,the search engines Baidu,Bing,and 360 were selected as test objects to evaluate their performance in various aspects.The test results were reflected by the anomaly rate and average Jaccard coefficient,which will provide a reference for search engine users in selecting the appropriate search engine,and also help search engine developers to find and remove program errors and improve existing software design.