Some embodiments provide a non-transitory machine-readable medium that stores a program. The program parses a knowledge graph comprising a plurality of nodes and a plurality of edges to identify metadata associated with the plurality of nodes and the plurality of edges. Each edge in the plurality of edges connects a pair of nodes in the plurality of nodes. Each node in the plurality of nodes represents an entity. Each edge in the plurality of edges connecting a particular pair of nodes in the plurality of nodes represents a relationship between two entities represented by the particular pair of nodes. The program further determines a set of token definitions. The program also generates a set of syntax rules based on the identified metadata and the set of token definitions.