Purify noisy data from annotated images using Montylingua and control redundant term

Dynamic growths in the field of digital data and new techniques (manual and automatic) are introduced to tag images. Tagging of an object within the image is labeled in different terms base on the user perception. LabelMe is the image datasets that give a user online access to labeled object through by webtool. However, there are a number of noisy terms and errors found in the annotated list. Nevertheless, sometime a user tags the same objects with repeated terms. It requires to pruning the dataset from errors, noisy keywords and reduces to one instance term. This paper uses Montylingua for two purposes. First, it converts the tag term into base form. Second it purifies the irrelevant terms from the list. Next reduce the repeated terms into one instance and display their total count of occurrence. An experiment work, it shows that the purified list of the tagging has successfully removed from the annotated images. The result depicts through tagging ratio as well as degree of retrieval for effective achieved.