The Keyword Suggester is a semi-automatic image tagging system that can be applied to external image URL or to internal image IDs. The suggestions are calculated based on the keywords of indexed images that are visually similar to the input image.
There are many parameters to control the suggestion process. However, the Keyword Suggester can only produce meaningful suggestions if there are visually similar images with meaningful keywords in your collection. The keywords within your index act as dictionary for the Keyword Suggester. Suggestions are always made from existing keywords. Therefore, if existing keywords are of poor quality the suggestions will suffer. No meaningful suggestions will be returned if there are no visually similar images with a similar context available in your current collection.
The Keyword Suggester assists you to find good keywords from your keywords and categories for new images. It furthermore helps you to keep keywords and categories for new images in the same wording style because the suggestions are calculated based on your individual keyword dictionary. Take a look at the ksuggest parameter and the response format in Get Keyword Suggestions for a usage and response format example.
The Keyword Suggester is designed as a semi-automatic tool and works best with user interactions to iteratively retrieve optimal suggestions. To get the most out of the Keyword Suggester you should implement a refinement workflow (see Refinement process).
All of the following parameters can be configured globally in
solrconfig.xml (see Configuration) or at runtime query parameters..