Peculiarities of Avestan Manuscripts for Computational Linguistics

This paper will discuss several computational tools for creating a stemma of Avestan manuscripts, such as: a letter similarity matrix, a morphological expander, and co-occurrence networks. After a short introduction to Avestan and Avestan manuscripts and a representation of Avestan peculiarities concerning the creation of stemmata, the operatability of the above-mentioned tools for this text corpus will be discussed. Finally, I will give a brief outlook on the complexity of a database structure for Avestan texts. Introduction The Avesta, represented by the edition of GELDNER (1886-96), appears to be a sort of Bible containing several books or chapters, cf. SKJÆRVØ’s “sacred book of the Zoroastrians” (2009: 44); and, indeed, in Middle Iranian times (i.e., before 600 AD) there existed a kind of text corpus, rather than ‘a book’, of holy texts (CANTERA 2004). However, GELDNER’s edition disguises the actual texts of the manuscripts because what we have today is not a book but a collection of ceremonies attested in various manuscripts. Avestan is the term for an Old Iranian language, as such a member of the IndoEuropean language family. The actual name of the language is not known to us. The name ‘Avestan’ is taken from Middle Persian texts which refer to their religious text corpus as the “abestā(g)”. When manuscripts containing these religious texts came to light for European research, they were referred to as “Avesta” and the language as “Avestan”. Avestan is known to us in two varieties, called “Old Avestan” and “Young Avestan”. This is so because they display two different chronological layers of Avestan. However, they also differ in some linguistic respect so that they represent two different dialects of the same language (e.g., genitive singular of xratu“wisdom” is xratə̄uš in Old Avestan but xraθβō in Young Avestan, for further examples see DE VAAN 2003: 8ff.). The Avestan manuscripts (henceforth MS) can be sorted into several groups, the main grouping is: 1) the ‘Pahlavi-MSs’, and 2) the ‘Sade-MSs’. The Pahlavi-MSs contain the Avestan text plus its translation and commentaries, generally Middle Persian, but there are translations into Sanskrit, Gujarati and/or New Persian as well. The Sade-MSs (i.e., the “pure” MS) only contain ritual instructions in Middle Persian, etc., besides the Avestan text. The Pahlavi-MS served as exegetical texts written for scholarly use only. On the contrary, the Sade-MSs were for the daily use in the ceremonies. These different purposes had an influence on the copying process (cf. Section 1). The aforementioned grouping can be made by first glance at the MS because of the various writings these MSs do or do not contain. Besides the grouping into Pahlaviand Sade-MSs, the MSs are further classified into different ceremonies. There are four of them: the Yasna Rapihwin, Vīsprad, Yašt, and Vīdēvdād ceremony. Depending on the season or on the deity who is invoked, there are further differences in what is otherwise the same