Flag format for irish ispell dictionaries
Irish dictionaries for Ispell(1) supports 3 prefix and 14 suffix flags. For a detailed description of how Ispell handles flags and capitalization, see Ispell(4). This manual page only describes flags usable in dictionaries built using the gaeilge.aff affix file.
The definitive meanings of the affixes is given in the file /usr/lib/ispell/gaeilge.aff but some of the reasoning and notes are given here.
Irish has an unusual (unique?) capitalization rule, which needs special handling. Prefixes for eclipsis are not capitalized., e.g. an tEarrach. Ispell(5) assumes that in an uncapitalized word, the first letter may be capitalized, so Earrach would be incorrectly prefixed and capitalized as Tearrach.
The file irish-prefixes.pl is a perl(1) script to preprocess the wordlist, and add 'by hand' the correct capitalization. It will turn
Earrach/P
into
Earrach
tEarrach etc. It strips off the E and P flags used for eclipsis and leaves any other flags intact. The word should be Capitalized in the the word list; if the uncapitalized version is left in the word list with the flag, it will generate incorrect words, so enter the uncapitalized and capitalized versions seperately.
One 'ordinary' prefix is handled by Ispell, H is used for aspiration of nouns, etc. (Verbs are handled separately).
Twelve suffixes are used for handling regular verbs. We use three suffixes each per conguation: one listing all suffixes available to both aspirated and unaspirated forms, one listing aspirated forms only, and one for unaspirated forms. So, for the verb mol for example, we have in the words list
mol/BC
mhol/BD
so the 'B' set of endings apply to both forms, the 'C' set to unaspirated only, and 'D' to aspirated only. This pattern is repeated for all four conjugations.
Irregular verbs just appear in the list in toto.
Two other suffixes are currently defined, , 'O' for Caolú (attenuation), and 'Q'for Leathnú (broadening).
The dictionary is very small, and pretty incomplete. Any help would be appreciated. Please mail additions and corrections to [email protected]
Ispell(1), Ispell(5), english(5)