MORPHOLOGICAL ANALYSIS OF HISTORICAL LANGUAGES

This paper describes the design of a system for parsing ancient Greek morphologically, taking into account the distinctive requirements of studying a historical language. Sets of stems and inflectional patterns identified by URNs are expressed in a specified orthography, and are loaded dynamically when the parser is built. Analytical results are also expressed with URN values for lexical entity and form, so that replies identify both the result of the analysis and the logic leading to the result. The underlying engine uses finite transducers to collect possible analyses of Greek words stripped of accents. A wrapper program implements an analysis by synthesis' algorithm by adding appropriate accents for each candidate analysis, and comparing the result to the originally submitted form. Parsers can be built or tailored to specific corpora without programming, by editing tables of stems and inflectional patterns.