Measuring the relevance of retrieved images in search engines based on Persian language writing styles

The aim of this study was to investigate the image retrieval from selected search engines according to the written and semantic features of Persian language and determine their relevance using recall and precision formulas and to identify the most efficient search engine in retrieving images in Persian and by survey-analytical method. It was done using direct observation technique. After reviewing related researches, search keywords list was formed in the form of a checklist based on the written and semantic features of Persian language. Each of these keywords in the studied search engines, including two general search engines Google and Bing and Duckduckgo semantic search engine, which are among the most used search engines and have also provided the ability to search for images in Persian, search and the number of relevant and unrelated retrieved results were recorded. Then, the recall and precision of search results in each search engine were calculated and the relevance of images based on these features in each of the studied search engines was investigated. A variety of descriptive statistical techniques were applied to analyze the data along with Kolmogorov-Smirnov, Shapiro-Wilk, Kruskal-Wallis and Friedman tests. Findings demonstrated that Google, Bing and Duckduckgo search engines do not pay enough attention to the written and semantic features of Persian language and many of these features are ignored while searching and retrieving images. In the present study, Google search engine had a higher recall and precision than the other two search engines, and despite the claim of semantic search engines to provide better and more relevant information than other search engines, Duckduckgo search engine did not show good performance in retrieving images related to the written and semantic part of Persian language. There is also a significant difference between the recall and the precision of the three studied search engines.