Models for natural language processing use statistics to collect a wealth of information about word meanings. In "Throug