Search Strategy Effectiveness and Relevance of YouTube Videos

摘要

Precision and recall are two widely accepted measures toInformation Retrieval (IR) system performance. Differentsearch strategies can yield different precisions and recallsfor the same IR system, search engine or website.Therefore, precision and recall can also be used to measuresearch strategy effectiveness. YouTube® is a popular socialmedia website which hosts the largest number of usergeneratedvideos on the web. The YouTube® search engineonly provides text words search function and search resultsare retrieved through the system matching these searchterms to video descriptions, tags, comments, etc. Therefore,precision and recall can also be used to measure searchstrategy effectiveness. This paper aims to test if a searchstrategy using multiple search terms is effective byanalyzing the relevance of the retrieved YouTube® videoson Smokeless Tobacco Products (STP) and YouTube®provided video usage statistics and community engagementstatistics.17 search terms were used to retrieve YouTube® videos onSTP which were recorded in a master file. All duplicatevideos from the retrieved results were excluded. Uniquevideos were then selected based on pre-defined inclusionand exclusion criteria set by the research team. A sample of440 unique videos was randomly selected for data analysis.SPSS 19.0 was used for data analysis. Descriptive analysis,precisions and recalls, logistic regression, and odds ratiotest results are reported. The study found that only a fewsearch terms out of the total 17 terms are more effective interms of relevance of the retrieved videos. Also, YouTube®provided video usage statistics and community engagementstatistics did not have significant association with therelevance of the retrieved videos. Recommendations forpotential solutions on designing more effective searchstrategies are provided.