The $q parameter is used to access a special full-text index that searches within the dataset. The full-text index spans all of the fields of the dataset, so think of it more like using a search engine than performing a SQL query.
For example, to search for the string “Islands” inside our earthquakes dataset:
https://soda.demo.socrata.com/resource/4tka-6guv.json?$q=Islands
Multiple search terms are ANDed together, rather than used as a phrase search, and words are stemmed before indexing and before each query. Stemming reduces each word down to its base or root form, so plural nouns like “criminals” will be reduced down to the singular “criminal”, and verbs like “stealing” will reduced down to their root of “steal”. So, searching with $q=stealing bicycles would match on the phrase “he was looking for a bicycle to steal” as well as the words “steals” and “bicycles” in different columns of a given record.