A search server stores a plurality of keywords, each associated with an entity type. The server stores grammar rules, each specifying one or more entity types. The server receives a search query with search terms and generates tokens by splitting the search terms, each token associated with a token position that indicates a location of the token within the query. The server generates n-grams. Each n-gram is associated with a start token position and an end token position. The server determines entity types for n-grams that correspond to the keywords and generates a mapping that maps the received entity types and the start token positions of the n-grams that correspond with the received entity types to the end token positions of the n-grams that correspond with the received entity types. The server identifies grammar rules based on the mapping and transmits a result based on the identified grammar rules.