# Lithuanian letter names that espeak-ng gets wrong when a word is spelled out.
#
# letter <TAB> IPA <TAB> word prefixes after which it is an abbreviation, not a
# letter (comma-separated, optional). Lines starting with # are ignored.
#
# el: espeak-ng's Lithuanian dictionary expands "el" to "elektroninis", because
#     "el." is the common abbreviation for "el. paštas" (e-mail). Right for the
#     abbreviation, wrong for the letter name of L, which appears in every
#     spelled-out abbreviation containing it (LRT, MTL, LT). A trailing period
#     cannot tell the two apart (at the end of a sentence the letter carries
#     one too), so the abbreviation is recognized by the word that follows.
#     The IPA matches the corpus' own inventory: espeak renders a final
#     Lithuanian "l" as the syllabic l̩ used throughout the training data.
# i:  lengthened because as the last letter of a spelled-out abbreviation
#     ("VMI") a short i is simply not heard; in the corpus the letter name
#     appears mid-sentence, where it is short.
# z, ž: espeak-ng gives both the same name, ʑˈee, so the initials of "Zigmas"
#     and "Žygimantas" are read identically. It also names Š as "eš" (ˋeʃ),
#     the pattern of ef/el/em/en/er/es, but Lithuanian Š is "šė". The forms
#     below are the corpus' own: žė ʒˈee (5), zė zˈee (2), šė ɕˈee (1),
#     against vė vʲˈee (44) and bė bʲˈee (22), which espeak already gets right.
#     Y is left alone on the same grounds: espeak reads it "igrek" and so does
#     the corpus (ˌiːɡrʲek), even though the school name is "i ilgoji".
el	ˈel̩	pašt
i	ˈiː
z	zˈee
ž	ʒˈee
š	ɕˈee