This paper describes the natural language processing component of a speech understanding system. The CHRONUS system was designed for the ATIS (Air Travel Information Service) task 1. The overall project consists of building a speech understanding system within the domain of airline reservations. Ideally, a user communicates by speaking to a computer to inquire about flights, fares, and other travel information. This task differs drastically from the previous speech understanding projects because the input to the system is spontaneous speech and the vocabulary is unlimited, even though the domain is restricted. In the framework of our system, laid out in Pieraccini et al. [1, 2], the natural language processing front end to a relational database is described. The original aspect of this approach is that statistical-based algorithms are combined with linguistic knowledge to achieve the understanding task.