updated german abbreviations

[maia]

I suggest to use the following abbreviations array, which is based on the old one minus some abbreviations I don't think are used commonly, plus frequent abbreviations in a subset of my tweets database.

  ABBREVIATIONS = ['a', 'a.d', 'a.k.a', 'a.s.a.p', 'abg', 'alt', 'apr', 'art', 'aug', 'b',
                   'b.a', 'b.s', 'best', 'bgm', 'bldg', 'btw', 'buchst', 'bzgl', 'bzw', 'c',
                   'ca', 'co', 'd', 'd.d', 'd.h', 'd.r', 'dergl', 'dez', 'dgl', 'dr', 'dr ',
                   'dt', 'dzt', 'e', 'e.l', 'e.u', 'e.v', 'ehem', 'eig', 'etc', 'etc.p.p',
                   'eu', 'europ', 'ev', 'ev ', 'evtl', 'f', 'f.d', 'feat', 'feb', 'ff',
                   'fr', 'frz', 'ft', 'g', 'gg', 'ggf', 'ggü', 'griech', 'h', 'h.c', 'h.p',
                   'hon', 'hosp', 'hr', 'i', 'i.a', 'i.d', 'i.d.r', 'i.f', 'i.p', 'i.z',
                   'ii', 'iii', 'inkl', 'int', 'iv', 'ix', 'j', 'jan', 'jul', 'jun', 'k',
                   'k.a', 'k.i.z', 'k.o', 'k.u.k', 'kath ', 'l', 'l.a', 'lfd', 'lt', 'ltd',
                   'm', 'm.e', 'm.w', 'mag', 'max', 'me', 'med', 'mind', 'mio', 'mme', 'mr',
                   'mrd', 'mrs', 'ms', 'mwst', 'mär', 'n', 'nov', 'nr', 'o', 'o.k', 'o.ä',
                   'oct', 'okt', 'omg', 'oö', 'p', 'p.a', 'p.m', 'p.s', 'p.t', 'pol', 'pp',
                   'prof', 'präs', 'q', 'r', 'r.i.p', 'r.r', 'ranz', 'rd', 'rep', 'rt',
                   'russ', 's', 's.g', 'sen', 'sep', 'sog', 'st', 'std', 'str', 't', 'türk',
                   'u', 'u.a', 'u.a  ', 'u.a.m', 'u.a.v', 'u.k', 'u.s', 'u.s.w', 'u.u',
                   'u.v.a', 'u.v.m', 'u.ä', 'ungar', 'usf', 'usw', 'uvm', 'v', 'v.a', 'v.d',
                   'v.m', 'vgl', 'vi', 'vii', 'viii', 'vs', 'w', 'wg', 'wr', 'x', 'xi',
                   'xii', 'xiii', 'xiv', 'xix', 'xv', 'xvi', 'xvii', 'xviii', 'xx', 'y',
                   'z', 'z.b', 'z.t', 'z.z', 'z.zt', 'zb', 'zt', 'zw', 'zzt', 'ä', 'ö',
                   'öffentl', 'öst', 'österr', 'ü'].freeze

In case this array is too long, I can query my complete (but still non-representative) db of tweets and remove the ones with the least occurrence frequency.

[maia]

P.S. The above list contains all single letters, as each of these can be an abbreviated first name.

[diasks2]

Thanks! Added these in: https://github.com/diasks2/pragmatic_tokenizer/commit/0a8ecfdae1e87a632c8f99d81b998011b282525f

[maia]

Thanks!