Generating and Parsing Classical Greek

Abstract

This paper will describe the development of a rule based system for the analysis of Greek morphology and provide background for a Greek morphological database created by this system. While this paper focuses on classical Greek, a highly inflected Indo-European language, our work was intended from the outset to provide a case study that would illustrate the problems of morphological analysis and the possibilities that a working morphological system might open up for research M´uch of the impetus for out work on Greek morphological sprang from work in Akkadian and Sumerian, and from the frustrations that scholars inevitably face when they attempt to work with primary material beyond their immediate area of specialization. A system for morphological analysis is both a tool for those specializing in the target language and a gateway whereby those from other disciplines may enter more deeply the primary soruce texts of ancient Greece.

Working with a database of 40,000 stems, 13,000 inflections, and 2,500 irregular forms, Morpheus, the parser described in this paper has (as of October 1990) been used to analyze roughly 3,000,000 words, including texts in a variety of dialects from the eighth century BC to the second century AD and has been used to create a morphological database. This morphological database plays a major role in binding together 40 megabytes of primary sources within the hypermedia database on ancient Greece developed by the Perseus Project.¹ At the same time, the database exists as a separate entity distinct from the larger Perseus database, and we plan in early 1992 to distribute for linguistic analysis a fuller version of the database, covering those Greek texts in the Thesaurus Linguae Graecae from Homer to the fourth century

This paper deals with the reasons why this parser was developed and the problems that we encountered. A parser for Greek morphology was not new even in 1984 when we began, but out parser was designed to control phenomena (such as Greek diacritics and dialects) that earlier work had not addressed. This paper will describe the initial purposes and rationale for devising our parser, and the problems that we encountered in striving for a higher level of precision. Its purpose is to predict problems and highlight possible solutions for those working in Greek and to serve as an example of one approach for those working in other languages.