Understanding LOLITA: Experiences in Teaching Large Scale Functional Programming

LOLITA is a large scale natural processing system written in the functional language Haskell. It consists of over 47,000 lines of code written over 180 dierent modules. There are currently 20 people working on the system, most of whom are Ph.D. students. The majority of research projects involve the development of an application which is written around a semantic network; the knowledge representation structure at the core of the system. Because of the type of various applications, developers often join the team with little or no functional programming experience. For this reason the task of teaching these developers to the level required to implement their respective applications, requires teaching at various levels of abstraction. The strategy chosen means that each researcher only needs to be taught at the particular level of abstraction at which they work. These abstractions give rise to the notion of a domain specic sublanguage; that is a programming style in which a dierent language is created for each desired level of abstraction. In this paper we show how functional languages provide the necessary framework to enable these sublanguages to be created.