Run sentiment analysis perpetually in the backend and just return results on request instead of running when requested.