soudabeh derakhshandeh; Fereshteh Sepehr; zahra abazari; neshaneh neshaneh
Abstract
Background and Objectives: Image indexing based on search engines in retrieving text-based and content-based indexed images using the Delphi technique.
Methodology: This research is applied in terms of classifying research according to the method of data collection (research design), using the Delphi ...
Read More
Background and Objectives: Image indexing based on search engines in retrieving text-based and content-based indexed images using the Delphi technique.
Methodology: This research is applied in terms of classifying research according to the method of data collection (research design), using the Delphi technique, and in terms of classifying research according to the purpose. The statistical population of the qualitative stage of this research included all the specialists who worked in the offices of national newspapers in Tehran and had sufficient aristocracy and mastery in all five search engines studied. Most of them, depending on the type of activity and field of work, work with a maximum of one or two search engines and are aware. Therefore, only 16 specialists were selected as the available sample for Delphi panel members. In the present study, the experts reached a general consensus with twenty questions in four stages, which were indexed based on a range of five Likert options (from very weak to very good). Given the low Kendall coefficient in the fourth round, the low agreement of the panel members, and the significance of the third round, it is concluded that there was no increase in agreement in this round, and the polling process should be stopped. In this study, after reviewing the existing texts and sources by documentary method (library study), 150 specialized questions in the field were collected. After discussion and exchange of views with experts and professors in this field, 31 questions were approved for implementation, which with 3 main components separately [image search engine evaluation criteria (nineteen questions), image retrieval based on text-based indexing (six questions), And image retrieval was performed based on content-based indexing (six questions). In the meantime, Hamshahri newspaper, with its various publications as well as provincial special issues, magazines (My Land, 24, Children, Health, Youth, Stories, Knowledge, Clues, Story Books, Advertising Brochures, Exhibition Brochures, etc.) [In addition to the 5 selected image search engines, there were other very good image search engines in this field that even ranked very well on Alexa (the international website for ranking sites and blogs); However, due to the lack of use of this language and inefficiency in Iran and the lack of a specific audience, we have ignored their choice in this study]. The results of 9 experts showed that the questionnaire's relative validity coefficient of 32 items out of 40 items was higher than the critical coefficient value of 0.78. However, the relative validity of the eight-item content was less than the critical coefficient and was omitted. Therefore, in the relative coefficient index, the content validity of 32 questionnaire items was confirmed. Also, the content validity index of the other 31 questionnaire items was higher than the standard value of 0.79. As a result, 31 questionnaire items were approved in terms of two relative content validity coefficients. In addition, Cronbach's alpha coefficient for the reliability of the questionnaire with 31 items on 16 experts showed that it is equal to 0.916, which is a high coefficient.
Findings: Google's search engine showed a higher image retrieval rate based on their evaluation criteria. There was no significant difference between the studied search engines from the perspective of experts in the field of image retrieval based on text-based indexing at the level of P <0.05. Yandex search engine has a higher content indexing based on indexing based on more content at the level of P <0.05. Also, the Google search engine is significantly more efficient at the level of P \u003c0.05 in terms of retrieving images based on the areas under study. The results of the present study indicate that: 1. In the search engines, from the point of view of experts in the field of image retrieval based on text-based indexing, there are more or fewer differences, so the highest average, in this case, belonged to the Google search engine, and the lowest average belonged to the Pinterest search engine. 2. Experts in image retrieval based on content-based indexing see some differences between the search engines, so the Yandex search engine showed the highest average in this case, and the Yahoo search engine had the lowest average among the surveyed search engines. 3. Regarding image retrieval based on evaluation criteria, there are some differences between the average search engines from the perspective of experts. Hence, the average Google search engine is higher than the average of other search engines, while the Yandex search engine, in this case, is the lowest. Has had an average. 4. From the experts' point of view, there are differences between the studied search engines regarding the most efficient search engine in retrieving images based on the researched areas. The results show that Google's search engine has a much higher average than other studied search engines and the lowest average. Has been to the Pinterest search engine.
Discussion: Google's general search engines perform better than other search engines (Yahoo, Bing, Pinterest, and Yandex) in retrieving images; web image searchers can also choose the search engine that suits their needs and interior designers for better design. In addition, these results can be generalized to similar areas, and search engine designers will find out which indexing method to use to retrieve images better. In conclusion, it is suggested that to pay more attention to the indexing and retrieval of images in search engines; designers should consider such features based on the main components identified separately: A. Criteria for image search engines (add a special code for image copyright, apply all components of this research as a menu and submenu in image search engines, notify the owners of the email (images) when uploading images for support) B. Consider image retrieval based on content-based indexing (possibility to combine multiple colors simultaneously, searchable image edges to be able to draw shapes, simultaneous searchability of image content information).
