In this project, we process each user query Q using the learned models from our previous work in order to 1) extract all of the query components that convey structural information, and 2) identify the requisite message information that should be conveyed by the desired graphic. In addition, we form a bag of words Qt consisting of all words in query Q.