A computer provides search suggestions by capturing data from a first application, wherein the data includes data displayed on a screen by the first application. The computer segments the data from the first application into blocks. The computer extracts words from the data from the first application. The computer generates tokens of phrases and words using natural language processing tokenization. The computer scores the blocks based upon the data from the first application. The computer scores the tokens based upon the scoring of the blocks and the data from the first application. The computer detects a switch action from the first application to a second application, wherein the second application contains a search function. The computer provides the scored tokens to the search function of the second application.