Importantly, Ramani says the search is a multistep process, allowing users to refine their results several times and allowing the search tool to learn what the user wants; this relevance feedback requires neural network technology that bridges the gap between the user's perception of the part and what is actually in the database.