摘要：With the tremendous booming of uploading and sharing images over the web, there is an es-sential demand for illustrating the content of such images. Title and description provide a po-etic explanation for the images while tags tend to be more appropriate in terms of determining the content. This refers to the keywords obtained from tags which facilitate the process of fig-uring out the exact content. Several methods have been implemented toward enhancing the tag suggestion. However, there are cases where the suggestion misleads the search. Such cases are targeting a domain-specific for instance, searching for an image of the house using sign lan-guage, most of the obtained results will be full of house images. This is due to overlapping between two domains of specific which are Houses and Sign language. Therefore, this study proposed a domain-specific algorithm for Arabic tag suggestions. The data is a set of images of objects customized for sign language which have been collected from an institution for the disabled. Hence, using a game-based approach those images will be tagged. Consequently, those tags will undergo preprocessing including translation, normalization, and tokenization. After that, a domain-specific algorithm will be carried out using three similarity measures which are Cosine, Dice, and Jaccard. The evaluation has been performed using the common information retrieval metrics Precision, Recall, and F-measure.
