Automatic analysis of the Old Babylonian finite verb

Aleksi Sahala 2010
University of Helsinki

Abstract

General

The goal of my thesis was to write a regular expression based program (in Python 2.5) that was able to analyze Akkadian nonlinear morphology. Main focus was set on finite verbs as they carry all characteristics of the Akkadian grammar. Program was written to support G, D, Š and N-stems of strong, I-nun, I-waw, I-aleph, I-yod, II-infirmae, III-infirmae and quadriliteral verbs in present, perfect, preterite, imperative and stative. All personal suffixes, dative, accusative, ventive and consecutive are supported.

The program uses customizable lexicon and affix files (XML) which allow user to add possibly missing variants. The lexicon includes most of the attested Akkadian verbal roots (1307) based on Concise Dictionary of Akkadian (CDA).

Inputs can be given in transliteration or transcription (preferred). Distinctive vowel length may be ignored to overcome some problems concerning transliteration.

Parser

Analysis process is divided into two phases: Affix analysis and root analysis. In the affix analysis the input is compared with all affixes listed in the grammar files. Found affixes are cut off and added into a stack. The deaffixed input (stem) is then indentified by comparing it to regular expression stem patterns; eg. pattern R1-a-R2*2-V1-R3 stands for strong G-present parras, şabbit etc.

After affixes and stem have been identified, radicals (R1-R3) and opaque root vowel (V1 or V2) are picked and validated by comparing them with the lexicon. Should the analyzed input be a valid finite verb, the program will generate a morphological gloss for it and print out an analysis.

A small scale validation (100 random samples from Enūma Eliš) indicated the program having an error percent of 4-12% depending on input form. Transcription with vowel quantity ignored (!ilv = OFF) gave the best results.