Phonological changes between Old and Middle English (an algorithmic approach)

Abstract

C lem t eto bakal a rsk e pr ace bylo vytvo rit automatick y analyz ator diachronn ch fonetick ych zm en v angli ctin e, kter y je konkr etn e zam e ren na zm eny mezi staroangli ctinou a st redn angli ctinou. Vstupem programu jsou dv e slova: jedno je staroanglick e a druh e je jeho pozd ej s verz ve st redn angli ctin e. Vzhledem k tomu, ze tyto zm eny prob haj na mluven e form e jazyka, vypl yv a n am ze zad an n ekolik d l c ch ukol u: vytvo rit automatick y p revad e c ze staroanglick e psan e formy do odpov daj c ho p repisu v yslovnosti, tot e z pro st redn angli ctinu (co z je ale d ky povaze tehdej s ho pravopisu strukturn e slo zit ej s ) a nakonec algoritmus, kter y mezi navr zen ymi v yslovnostmi pro dan a obdob nalezne cestu ve form e pravd epodobn e sekvence fonetick ych zm en, kter a by vysv etlovala vztah mezi dv ema verzemi slova, kter e jsme dostali na vstupu. Ve fonetick ych zm en ach panuje ur cit a pravidelnost a tud z je lze z velk e c asti popsat pomoc sady pravidel. Tato pravidla jsou v t eto pr aci formalizov ana tak, aby bylo mo zno je pozd eji zkombinovat s funkcionalitou standardn ch informatick ych algoritm u. Reprezentace t echto pravidel vych az ze syntaxe regul arn ch v yraz u s ur cit ymi upravami a p ridan ymi funkcemi. Nejd ule zit ej s p ridanou funkcionalitou je zde...

The purpose of this thesis was to create an automatic analyzer of phonetic changes in the historical development of English, namely between Old and Middle English. The analyzer gets an OE and ME form of a word at the input and produces a suggestion of an explanatory sequence of changes connecting these two forms as the output. As sound changes operate mainly on the spoken language, three main subtasks emerge: to create a grapheme-to-phoneme convertor for Old English, the same (although structurally more complicated) for Middle English and a core algorithm that searches for possible sequences that would explain the development of the word between its OE and ME spoken forms. There is a certain regularity in the diachronic phonetic changes, and therefore these can be translated into a set of rules. A general framework is proposed and implemented that works with these rules in a certain formalized fashion that allows their integration within standard algorithms. The form of the rules is largely adopted from regular expressions, with some alterations and additions, the most important being the possibility of using wildcards that are based on phonetic properties. A database of phoneme representations and phonetic properties is taken from Kirshenbaum's 1992 proposal for ASCII representation of IPA symbols. In the...