well the intro line says "extremely fast probabilistic counting of item frequencies" further down, "Such memory vs. accuracy tradeoffs are sometimes desirable

well the intro line says "extremely fast probabilistic counting of item frequencies" further down, "Such memory vs. accuracy tradeoffs are sometimes desirable in NLP, where being able to handle very large collections is more important than whether an event occurs exactly 55,482x or 55,519x." if you are on github, perhaps raise an issue to add an example similar to yours as well