You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

607 lines
28 KiB
Python

#### PATTERN | DE | INFLECT ########################################################################
# -*- coding: utf-8 -*-
# Copyright (c) 2012 University of Antwerp, Belgium
# Author: Tom De Smedt <tom@organisms.be>
# License: BSD (see LICENSE.txt for details).
####################################################################################################
# Regular expressions-based rules for German word inflection:
# - pluralization and singularization of nouns and adjectives,
# - conjugation of verbs,
# - attributive and predicative of adjectives,
# - comparative and superlative of adjectives.
# Accuracy (measured on CELEX German morphology word forms):
# 75% for gender()
# 72% for pluralize()
# 84% for singularize() (for nominative)
# 87% for Verbs.find_lemma()
# 87% for Verbs.find_lexeme()
# 98% for predicative
from __future__ import unicode_literals
from __future__ import division
from builtins import str, bytes, dict, int
from builtins import map, zip, filter
from builtins import object, range
import os
import sys
import re
try:
MODULE = os.path.dirname(os.path.realpath(__file__))
except:
MODULE = ""
sys.path.insert(0, os.path.join(MODULE, "..", "..", "..", ".."))
from pattern.text import Verbs as _Verbs
from pattern.text import (
INFINITIVE, PRESENT, PAST, FUTURE,
FIRST, SECOND, THIRD,
SINGULAR, PLURAL, SG, PL,
INDICATIVE, IMPERATIVE, SUBJUNCTIVE,
PROGRESSIVE,
PARTICIPLE, GERUND
)
sys.path.pop(0)
VERB, NOUN, ADJECTIVE, ADVERB = "VB", "NN", "JJ", "RB"
VOWELS = "aeiouy"
re_vowel = re.compile(r"a|e|i|o|u|y", re.I)
is_vowel = lambda ch: ch in VOWELS
#### ARTICLE #######################################################################################
# German inflection of depends on gender, role and number + the determiner (if any).
# Inflection gender.
# Masculine is the most common, so it is the default for all functions.
MASCULINE, FEMININE, NEUTER, PLURAL = \
MALE, FEMALE, NEUTRAL, PLURAL = \
M, F, N, PL = "m", "f", "n", "p"
# Inflection role.
# - nom = subject, "Der Hund bellt" (the dog barks).
# - acc = object, "Das Mädchen küsst den Hund" (the girl kisses the dog).
# - dat = object (indirect), "Der Mann gibt einen Knochen zum Hund" (the man gives the dog a bone).
# - gen = property, "die Knochen des Hundes" (the dog's bone).
NOMINATIVE, ACCUSATIVE, DATIVE, GENITIVE = SUBJECT, OBJECT, INDIRECT, PROPERTY = \
"nominative", "accusative", "dative", "genitive"
article_definite = {
("m", "nom"): "der", ("f", "nom"): "die", ("n", "nom"): "das", ("p", "nom"): "die",
("m", "acc"): "den", ("f", "acc"): "die", ("n", "acc"): "das", ("p", "acc"): "die",
("m", "dat"): "dem", ("f", "dat"): "der", ("n", "dat"): "dem", ("p", "dat"): "den",
("m", "gen"): "des", ("f", "gen"): "der", ("n", "gen"): "des", ("p", "gen"): "der",
}
article_indefinite = {
("m", "nom"): "ein" , ("f", "nom"): "eine" , ("n", "nom"): "ein" , ("p", "nom"): "eine",
("m", "acc"): "einen", ("f", "acc"): "eine" , ("n", "acc"): "ein" , ("p", "acc"): "eine",
("m", "dat"): "einem", ("f", "dat"): "einer", ("n", "dat"): "einem", ("p", "dat"): "einen",
("m", "gen"): "eines", ("f", "gen"): "einer", ("n", "gen"): "eines", ("p", "gen"): "einer",
}
def definite_article(word, gender=MALE, role=SUBJECT):
""" Returns the definite article (der/die/das/die) for a given word.
"""
return article_definite.get((gender[:1].lower(), role[:3].lower()))
def indefinite_article(word, gender=MALE, role=SUBJECT):
""" Returns the indefinite article (ein) for a given word.
"""
return article_indefinite.get((gender[:1].lower(), role[:3].lower()))
DEFINITE = "definite"
INDEFINITE = "indefinite"
def article(word, function=INDEFINITE, gender=MALE, role=SUBJECT):
""" Returns the indefinite (ein) or definite (der/die/das/die) article for the given word.
"""
return function == DEFINITE \
and definite_article(word, gender, role) \
or indefinite_article(word, gender, role)
_article = article
def referenced(word, article=INDEFINITE, gender=MALE, role=SUBJECT):
""" Returns a string with the article + the word.
"""
return "%s %s" % (_article(word, article, gender, role), word)
#### GENDER #########################################################################################
gender_masculine = (
"ant", "ast", "ich", "ig", "ismus", "ling", "or", "us"
)
gender_feminine = (
"a", "anz", "ei", "enz", "heit", "ie", "ik", "in", "keit", "schaf", "sion", "sis",
"tät", "tion", "ung", "ur"
)
gender_neuter = (
"chen", "icht", "il", "it", "lein", "ma", "ment", "tel", "tum", "um", "al", "an", "ar",
"ät", "ent", "ett", "ier", "iv", "o", "on", "nis", "sal"
)
gender_majority_vote = {
MASCULINE: (
"ab", "af", "ag", "ak", "am", "an", "ar", "at", "au", "ch", "ck", "eb", "ef", "eg",
"el", "er", "es", "ex", "ff", "go", "hn", "hs", "ib", "if", "ig", "ir", "kt", "lf",
"li", "ll", "lm", "ls", "lt", "mi", "nd", "nk", "nn", "nt", "od", "of", "og", "or",
"pf", "ph", "pp", "ps", "rb", "rd", "rf", "rg", "ri", "rl", "rm", "rr", "rs", "rt",
"rz", "ss", "st", "tz", "ub", "uf", "ug", "uh", "un", "us", "ut", "xt", "zt"
),
FEMININE: (
"be", "ce", "da", "de", "dt", "ee", "ei", "et", "eu", "fe", "ft", "ge", "he", "hr",
"ht", "ia", "ie", "ik", "in", "it", "iz", "ka", "ke", "la", "le", "me", "na", "ne",
"ng", "nz", "on", "pe", "ra", "re", "se", "ta", "te", "ue", "ur", "ve", "ze"
),
NEUTER: (
"ad", "al", "as", "do", "ed", "eh", "em", "en", "hl", "id", "il", "im", "io", "is",
"iv", "ix", "ld", "lk", "lo", "lz", "ma", "md", "mm", "mt", "no", "ns", "ol", "om",
"op", "os", "ot", "pt", "rk", "rn", "ro", "to", "tt", "ul", "um", "uz"
)
}
def gender(word, pos=NOUN):
""" Returns the gender (MALE, FEMALE or NEUTRAL) for nouns (majority vote).
Returns None for words that are not nouns.
"""
w = word.lower()
if pos == NOUN:
# Default rules (baseline = 32%).
if w.endswith(gender_masculine):
return MASCULINE
if w.endswith(gender_feminine):
return FEMININE
if w.endswith(gender_neuter):
return NEUTER
# Majority vote.
for g in gender_majority_vote:
if w.endswith(gender_majority_vote[g]):
return g
#### PLURALIZE ######################################################################################
plural_inflections = [
("aal", "äle" ), ("aat", "aaten"), ("abe", "aben" ), ("ach", "ächer"), ("ade", "aden" ),
("age", "agen" ), ("ahn", "ahnen"), ("ahr", "ahre" ), ("akt", "akte" ), ("ale", "alen" ),
("ame", "amen" ), ("amt", "ämter"), ("ane", "anen" ), ("ang", "änge" ), ("ank", "änke" ),
("ann", "änner" ), ("ant", "anten"), ("aph", "aphen"), ("are", "aren" ), ("arn", "arne" ),
("ase", "asen" ), ("ate", "aten" ), ("att", "ätter"), ("atz", "ätze" ), ("aum", "äume" ),
("aus", "äuser" ), ("bad", "bäder"), ("bel", "bel" ), ("ben", "ben" ), ("ber", "ber" ),
("bot", "bote" ), ("che", "chen" ), ("chs", "chse" ), ("cke", "cken" ), ("del", "del" ),
("den", "den" ), ("der", "der" ), ("ebe", "ebe" ), ("ede", "eden" ), ("ehl", "ehle" ),
("ehr", "ehr" ), ("eil", "eile" ), ("eim", "eime" ), ("eis", "eise" ), ("eit", "eit" ),
("ekt", "ekte" ), ("eld", "elder"), ("ell", "elle" ), ("ene", "enen" ), ("enz", "enzen" ),
("erd", "erde" ), ("ere", "eren" ), ("erk", "erke" ), ("ern", "erne" ), ("ert", "erte" ),
("ese", "esen" ), ("ess", "esse" ), ("est", "este" ), ("etz", "etze" ), ("eug", "euge" ),
("eur", "eure" ), ("fel", "fel" ), ("fen", "fen" ), ("fer", "fer" ), ("ffe", "ffen" ),
("gel", "gel" ), ("gen", "gen" ), ("ger", "ger" ), ("gie", "gie" ), ("hen", "hen" ),
("her", "her" ), ("hie", "hien" ), ("hle", "hlen" ), ("hme", "hmen" ), ("hne", "hnen" ),
("hof", "höfe" ), ("hre", "hren" ), ("hrt", "hrten"), ("hse", "hsen" ), ("hte", "hten" ),
("ich", "iche" ), ("ick", "icke" ), ("ide", "iden" ), ("ieb", "iebe" ), ("ief", "iefe" ),
("ieg", "iege" ), ("iel", "iele" ), ("ien", "ium" ), ("iet", "iete" ), ("ife", "ifen" ),
("iff", "iffe" ), ("ift", "iften"), ("ige", "igen" ), ("ika", "ikum" ), ("ild", "ilder" ),
("ilm", "ilme" ), ("ine", "inen" ), ("ing", "inge" ), ("ion", "ionen"), ("ise", "isen" ),
("iss", "isse" ), ("ist", "isten"), ("ite", "iten" ), ("itt", "itte" ), ("itz", "itze" ),
("ium", "ium" ), ("kel", "kel" ), ("ken", "ken" ), ("ker", "ker" ), ("lag", "läge" ),
("lan", "läne" ), ("lar", "lare" ), ("lei", "leien"), ("len", "len" ), ("ler", "ler" ),
("lge", "lgen" ), ("lie", "lien" ), ("lle", "llen" ), ("mel", "mel" ), ("mer", "mer" ),
("mme", "mmen" ), ("mpe", "mpen" ), ("mpf", "mpfe" ), ("mus", "mus" ), ("mut", "mut" ),
("nat", "nate" ), ("nde", "nden" ), ("nen", "nen" ), ("ner", "ner" ), ("nge", "ngen" ),
("nie", "nien" ), ("nis", "nisse"), ("nke", "nken" ), ("nkt", "nkte" ), ("nne", "nnen" ),
("nst", "nste" ), ("nte", "nten" ), ("nze", "nzen" ), ("ock", "öcke" ), ("ode", "oden" ),
("off", "offe" ), ("oge", "ogen" ), ("ohn", "öhne" ), ("ohr", "ohre" ), ("olz", "ölzer" ),
("one", "onen" ), ("oot", "oote" ), ("opf", "öpfe" ), ("ord", "orde" ), ("orm", "ormen" ),
("orn", "örner" ), ("ose", "osen" ), ("ote", "oten" ), ("pel", "pel" ), ("pen", "pen" ),
("per", "per" ), ("pie", "pien" ), ("ppe", "ppen" ), ("rag", "räge" ), ("rau", "raün" ),
("rbe", "rben" ), ("rde", "rden" ), ("rei", "reien"), ("rer", "rer" ), ("rie", "rien" ),
("rin", "rinnen"), ("rke", "rken" ), ("rot", "rote" ), ("rre", "rren" ), ("rte", "rten" ),
("ruf", "rufe" ), ("rzt", "rzte" ), ("sel", "sel" ), ("sen", "sen" ), ("ser", "ser" ),
("sie", "sien" ), ("sik", "sik" ), ("sse", "ssen" ), ("ste", "sten" ), ("tag", "tage" ),
("tel", "tel" ), ("ten", "ten" ), ("ter", "ter" ), ("tie", "tien" ), ("tin", "tinnen"),
("tiv", "tive" ), ("tor", "toren"), ("tte", "tten" ), ("tum", "tum" ), ("tur", "turen" ),
("tze", "tzen" ), ("ube", "uben" ), ("ude", "uden" ), ("ufe", "ufen" ), ("uge", "ugen" ),
("uhr", "uhren" ), ("ule", "ulen" ), ("ume", "umen" ), ("ung", "ungen"), ("use", "usen" ),
("uss", "üsse" ), ("ute", "uten" ), ("utz", "utz" ), ("ver", "ver" ), ("weg", "wege" ),
("zer", "zer" ), ("zug", "züge" ), ("ück", "ücke" )
]
def pluralize(word, pos=NOUN, gender=MALE, role=SUBJECT, custom={}):
""" Returns the plural of a given word.
The inflection is based on probability rather than gender and role.
"""
w = word.lower().capitalize()
if word in custom:
return custom[word]
if pos == NOUN:
for a, b in plural_inflections:
if w.endswith(a):
return w[:-len(a)] + b
# Default rules (baseline = 69%).
if w.startswith("ge"):
return w
if w.endswith("gie"):
return w
if w.endswith("e"):
return w + "n"
if w.endswith("ien"):
return w[:-2] + "um"
if w.endswith(("au", "ein", "eit", "er", "en", "el", "chen", "mus", "tät", "tik", "tum", "u")):
return w
if w.endswith(("ant", "ei", "enz", "ion", "ist", "or", "schaft", "tur", "ung")):
return w + "en"
if w.endswith("in"):
return w + "nen"
if w.endswith("nis"):
return w + "se"
if w.endswith(("eld", "ild", "ind")):
return w + "er"
if w.endswith("o"):
return w + "s"
if w.endswith("a"):
return w[:-1] + "en"
# Inflect common umlaut vowels: Kopf => Köpfe.
if w.endswith(("all", "and", "ang", "ank", "atz", "auf", "ock", "opf", "uch", "uss")):
umlaut = w[-3]
umlaut = umlaut.replace("a", "ä")
umlaut = umlaut.replace("o", "ö")
umlaut = umlaut.replace("u", "ü")
return w[:-3] + umlaut + w[-2:] + "e"
for a, b in (
("ag", "äge"),
("ann", "änner"),
("aum", "äume"),
("aus", "äuser"),
("zug", "züge")):
if w.endswith(a):
return w[:-len(a)] + b
return w + "e"
return w
#### SINGULARIZE ###################################################################################
singular_inflections = [
( "innen", "in" ), ( "täten", "tät"), ( "ahnen", "ahn"), ( "enten", "ent"), ( "räser", "ras"),
( "hrten", "hrt"), ( "ücher", "uch"), ( "örner", "orn"), ( "änder", "and"), ( "ürmer", "urm"),
( "ahlen", "ahl"), ( "uhren", "uhr"), ( "ätter", "att"), ( "suren", "sur"), ( "chten", "cht"),
( "kuren", "kur"), ( "erzen", "erz"), ( "güter", "gut"), ( "soren", "sor"), ( "änner", "ann"),
( "äuser", "aus"), ( "taten", "tat"), ( "isten", "ist"), ( "bäder", "bad"), ( "ämter", "amt"),
( "eiten", "eit"), ( "raten", "rat"), ( "ormen", "orm"), ( "ionen", "ion"), ( "nisse", "nis"),
( "ölzer", "olz"), ( "ungen", "ung"), ( "läser", "las"), ( "ächer", "ach"), ( "urten", "urt"),
( "enzen", "enz"), ( "aaten", "aat"), ( "aphen", "aph"), ( "öcher", "och"), ( "türen", "tür"),
( "sonen", "son"), ( "ühren", "ühr"), ( "ühner", "uhn"), ( "toren", "tor"), ( "örter", "ort"),
( "anten", "ant"), ( "räder", "rad"), ( "turen", "tur"), ( "äuler", "aul"), ( "änze", "anz"),
( "tten", "tte"), ( "mben", "mbe"), ( "ädte", "adt"), ( "llen", "lle"), ( "ysen", "yse"),
( "rben", "rbe"), ( "hsen", "hse"), ( "raün", "rau"), ( "rven", "rve"), ( "rken", "rke"),
( "ünge", "ung"), ( "üten", "üte"), ( "usen", "use"), ( "tien", "tie"), ( "läne", "lan"),
( "iben", "ibe"), ( "ifen", "ife"), ( "ssen", "sse"), ( "gien", "gie"), ( "eten", "ete"),
( "rden", "rde"), ( "öhne", "ohn"), ( "ärte", "art"), ( "ncen", "nce"), ( "ünde", "und"),
( "uben", "ube"), ( "lben", "lbe"), ( "üsse", "uss"), ( "agen", "age"), ( "räge", "rag"),
( "ogen", "oge"), ( "anen", "ane"), ( "sken", "ske"), ( "eden", "ede"), ( "össe", "oss"),
( "ürme", "urm"), ( "ggen", "gge"), ( "üren", "üre"), ( "nten", "nte"), ( "ühle", "ühl"),
( "änge", "ang"), ( "mmen", "mme"), ( "igen", "ige"), ( "nken", "nke"), ( "äcke", "ack"),
( "oden", "ode"), ( "oben", "obe"), ( "ähne", "ahn"), ( "änke", "ank"), ( "inen", "ine"),
( "seen", "see"), ( "äfte", "aft"), ( "ulen", "ule"), ( "äste", "ast"), ( "hren", "hre"),
( "öcke", "ock"), ( "aben", "abe"), ( "öpfe", "opf"), ( "ugen", "uge"), ( "lien", "lie"),
( "ände", "and"), ( "ücke", "ück"), ( "asen", "ase"), ( "aden", "ade"), ( "dien", "die"),
( "aren", "are"), ( "tzen", "tze"), ( "züge", "zug"), ( "üfte", "uft"), ( "hien", "hie"),
( "nden", "nde"), ( "älle", "all"), ( "hmen", "hme"), ( "ffen", "ffe"), ( "rmen", "rma"),
( "olen", "ole"), ( "sten", "ste"), ( "amen", "ame"), ( "höfe", "hof"), ( "üste", "ust"),
( "hnen", "hne"), ( "ähte", "aht"), ( "umen", "ume"), ( "nnen", "nne"), ( "alen", "ale"),
( "mpen", "mpe"), ( "mien", "mie"), ( "rten", "rte"), ( "rien", "rie"), ( "äute", "aut"),
( "uden", "ude"), ( "lgen", "lge"), ( "ngen", "nge"), ( "iden", "ide"), ( "ässe", "ass"),
( "osen", "ose"), ( "lken", "lke"), ( "eren", "ere"), ( "üche", "uch"), ( "lüge", "lug"),
( "hlen", "hle"), ( "isen", "ise"), ( "ären", "äre"), ( "töne", "ton"), ( "onen", "one"),
( "rnen", "rne"), ( "üsen", "üse"), ( "haün", "hau"), ( "pien", "pie"), ( "ihen", "ihe"),
( "ürfe", "urf"), ( "esen", "ese"), ( "ätze", "atz"), ( "sien", "sie"), ( "läge", "lag"),
( "iven", "ive"), ( "ämme", "amm"), ( "äufe", "auf"), ( "ppen", "ppe"), ( "enen", "ene"),
( "lfen", "lfe"), ( "äume", "aum"), ( "nien", "nie"), ( "unen", "une"), ( "cken", "cke"),
( "oten", "ote"), ( "mie", "mie"), ( "rie", "rie"), ( "sis", "sen"), ( "rin", "rin"),
( "ein", "ein"), ( "age", "age"), ( "ern", "ern"), ( "ber", "ber"), ( "ion", "ion"),
( "inn", "inn"), ( "ben", "ben"), ( "äse", "äse"), ( "eis", "eis"), ( "hme", "hme"),
( "iss", "iss"), ( "hen", "hen"), ( "fer", "fer"), ( "gie", "gie"), ( "fen", "fen"),
( "her", "her"), ( "ker", "ker"), ( "nie", "nie"), ( "mer", "mer"), ( "ler", "ler"),
( "men", "men"), ( "ass", "ass"), ( "ner", "ner"), ( "per", "per"), ( "rer", "rer"),
( "mus", "mus"), ( "abe", "abe"), ( "ter", "ter"), ( "ser", "ser"), ( "äle", "aal"),
( "hie", "hie"), ( "ger", "ger"), ( "tus", "tus"), ( "gen", "gen"), ( "ier", "ier"),
( "ver", "ver"), ( "zer", "zer"),
]
singular = {
"Löwen": "Löwe",
}
def singularize(word, pos=NOUN, gender=MALE, role=SUBJECT, custom={}):
""" Returns the singular of a given word.
The inflection is based on probability rather than gender and role.
"""
w = word.lower().capitalize()
if word in custom:
return custom[word]
if word in singular:
return singular[word]
if pos == NOUN:
for a, b in singular_inflections:
if w.endswith(a):
return w[:-len(a)] + b
# Default rule: strip known plural suffixes (baseline = 51%).
for suffix in ("nen", "en", "n", "e", "er", "s"):
if w.endswith(suffix):
w = w[:-len(suffix)]
break
# Corrections (these add about 1% accuracy):
if w.endswith(("rr", "rv", "nz")):
return w + "e"
return w
return w
#### VERB CONJUGATION ##############################################################################
# The verb table was trained on CELEX and contains the top 2000 most frequent verbs.
prefix_inseparable = (
"be", "emp", "ent", "er", "ge", "miss", "über", "unter", "ver", "voll", "wider", "zer"
)
prefix_separable = (
"ab", "an", "auf", "aus", "bei", "durch", "ein", "fort", "mit", "nach", "vor", "weg",
"zurück", "zusammen", "zu", "dabei", "daran", "da", "empor", "entgegen", "entlang",
"fehl", "fest", "gegenüber", "gleich", "herab", "heran", "herauf", "heraus", "herum",
"her", "hinweg", "hinzu", "hin", "los", "nieder", "statt", "umher", "um", "weg",
"weiter", "wieder", "zwischen"
) + ( # There are many more...
"dort", "fertig", "frei", "gut", "heim", "hoch", "klein", "klar", "nahe", "offen", "richtig"
)
prefixes = prefix_inseparable + prefix_separable
def encode_sz(s):
return s.replace("ß", "ss")
def decode_sz(s):
return s.replace("ss", "ß")
class Verbs(_Verbs):
def __init__(self):
_Verbs.__init__(self, os.path.join(MODULE, "de-verbs.txt"),
language = "de",
format = [0, 1, 2, 3, 4, 5, 8, 17, 18, 19, 20, 21, 24, 52, 54, 53, 55, 56, 58, 59, 67, 68, 70, 71],
default = {6: 4, 22: 20, 57: 55, 60: 58, 69: 67, 72: 70}
)
def find_lemma(self, verb):
""" Returns the base form of the given inflected verb, using a rule-based approach.
"""
v = verb.lower()
# Common prefixes: be-finden and emp-finden probably inflect like finden.
if not (v.startswith("ge") and v.endswith("t")): # Probably gerund.
for prefix in prefixes:
if v.startswith(prefix) and v[len(prefix):] in self.inflections:
return prefix + self.inflections[v[len(prefix):]]
# Common sufixes: setze nieder => niedersetzen.
b, suffix = " " in v and v.split()[:2] or (v, "")
# Infinitive -ln: trommeln.
if b.endswith(("ln", "rn")):
return b
# Lemmatize regular inflections.
for x in ("test", "est", "end", "ten", "tet", "en", "et", "te", "st", "e", "t"):
if b.endswith(x):
b = b[:-len(x)]; break
# Subjunctive: hielte => halten, schnitte => schneiden.
for x, y in (
("ieb", "eib"), ( "ied", "eid"), ( "ief", "auf" ), ( "ieg", "eig" ), ("iel", "alt"),
("ien", "ein"), ("iess", "ass"), ( "ieß", "" ), ( "iff", "eif" ), ("iss", "eiss"),
( "", "eiß"), ( "it", "eid"), ( "oss", "iess"), ( "öss", "iess")):
if b.endswith(x):
b = b[:-len(x)] + y; break
b = b.replace("eeiss", "eiss")
b = b.replace("eeid", "eit")
# Subjunctive: wechselte => wechseln
if not b.endswith(("e", "l")) and not (b.endswith("er") and len(b) >= 3 and not b[-3] in VOWELS):
b = b + "e"
# abknallst != abknalln => abknallen
if b.endswith(("hl", "ll", "ul", "eil")):
b = b + "e"
# Strip ge- from (likely) gerund:
if b.startswith("ge") and v.endswith("t"):
b = b[2:]
# Corrections (these add about 1.5% accuracy):
if b.endswith(("lnde", "rnde")):
b = b[:-3]
if b.endswith(("ae", "al", "öe", "üe")):
b = b.rstrip("e") + "te"
if b.endswith("äl"):
b = b + "e"
return suffix + b + "n"
def find_lexeme(self, verb):
""" For a regular verb (base form), returns the forms using a rule-based approach.
"""
v = verb.lower()
# Stem = infinitive minus -en, -ln, -rn.
b = b0 = re.sub("en$", "", re.sub("ln$", "l", re.sub("rn$", "r", v)))
# Split common prefixes.
x, x1, x2 = "", "", ""
for prefix in prefix_separable:
if v.startswith(prefix):
b, x = b[len(prefix):], prefix
x1 = (" " + x).rstrip()
x2 = x + "ge"
break
# Present tense 1sg and subjunctive -el: handeln => ich handle, du handlest.
pl = b.endswith("el") and b[:-2] + "l" or b
# Present tense 1pl -el: handeln => wir handeln
pw = v.endswith(("ln", "rn")) and v or b + "en"
# Present tense ending in -d or -t gets -e:
pr = b.endswith(("d", "t")) and b + "e" or b
# Present tense 2sg gets -st, unless stem ends with -s or -z.
p2 = pr.endswith(("s", "z")) and pr + "t" or pr + "st"
# Present participle: spiel + -end, arbeiten + -d:
pp = v.endswith(("en", "ln", "rn")) and v + "d" or v + "end"
# Past tense regular:
pt = encode_sz(pr) + "t"
# Past participle: haushalten => hausgehalten
ge = (v.startswith(prefix_inseparable) or b.endswith(("r", "t"))) and pt or "ge" + pt
ge = x and x + "ge" + pt or ge
# Present subjunctive: stem + -e, -est, -en, -et:
s1 = encode_sz(pl)
# Past subjunctive: past (usually with Umlaut) + -e, -est, -en, -et:
s2 = encode_sz(pt)
# Construct the lexeme:
lexeme = a = [
v,
pl + "e" + x1, p2 + x1, pr + "t" + x1, pw + x1, pr + "t" + x1, pp, # present
pt + "e" + x1, pt + "est" + x1, pt + "e" + x1, pt + "en" + x1, pt + "et" + x1, ge, # past
b + "e" + x1, pr + "t" + x1, x + pw, # imperative
s1 + "e" + x1, s1 + "est" + x1, s1 + "en" + x1, s1 + "et" + x1, # subjunctive I
s2 + "e" + x1, s2 + "est" + x1, s2 + "en" + x1, s2 + "et" + x1 # subjunctive II
]
# Encode Eszett (ß) and attempt to retrieve from the lexicon.
# Decode Eszett for present and imperative.
if encode_sz(v) in self:
a = self[encode_sz(v)]
a = [decode_sz(v) for v in a[:7]] + a[7:13] + [decode_sz(v) for v in a[13:20]] + a[20:]
# Since the lexicon does not contain imperative for all verbs, don't simply return it.
# Instead, update the rule-based lexeme with inflections from the lexicon.
return [a[i] or lexeme[i] for i in range(len(a))]
def tenses(self, verb, parse=True):
""" Returns a list of possible tenses for the given inflected verb.
"""
tenses = _Verbs.tenses(self, verb, parse)
if len(tenses) == 0:
# auswirkte => wirkte aus
for prefix in prefix_separable:
if verb.startswith(prefix):
tenses = _Verbs.tenses(self, verb[len(prefix):] + " " + prefix, parse)
break
return tenses
verbs = Verbs()
conjugate, lemma, lexeme, tenses = \
verbs.conjugate, verbs.lemma, verbs.lexeme, verbs.tenses
#### ATTRIBUTIVE & PREDICATIVE #####################################################################
# Strong inflection: no article.
adjectives_strong = {
("m", "nom"): "er", ("f", "nom"): "e" , ("n", "nom"): "es", ("p", "nom"): "e",
("m", "acc"): "en", ("f", "acc"): "e" , ("n", "acc"): "es", ("p", "acc"): "e",
("m", "dat"): "em", ("f", "dat"): "er", ("n", "dat"): "em", ("p", "dat"): "en",
("m", "gen"): "en", ("f", "gen"): "er", ("n", "gen"): "en", ("p", "gen"): "er",
}
# Mixed inflection: after indefinite article ein & kein and possessive determiners.
adjectives_mixed = {
("m", "nom"): "er", ("f", "nom"): "e" , ("n", "nom"): "es", ("p", "nom"): "en",
("m", "acc"): "en", ("f", "acc"): "e" , ("n", "acc"): "es", ("p", "acc"): "en",
("m", "dat"): "en", ("f", "dat"): "en", ("n", "dat"): "en", ("p", "dat"): "en",
("m", "gen"): "en", ("f", "gen"): "en", ("n", "gen"): "en", ("p", "gen"): "en",
}
# Weak inflection: after definite article.
adjectives_weak = {
("m", "nom"): "e", ("f", "nom"): "e" , ("n", "nom"): "e", ("p", "nom"): "en",
("m", "acc"): "en", ("f", "acc"): "e" , ("n", "acc"): "e", ("p", "acc"): "en",
("m", "dat"): "en", ("f", "dat"): "en", ("n", "dat"): "en", ("p", "dat"): "en",
("m", "gen"): "en", ("f", "gen"): "en", ("n", "gen"): "en", ("p", "gen"): "en",
}
# Uninflected + exceptions.
adjective_attributive = {
"etwas" : "etwas",
"genug" : "genug",
"viel" : "viel",
"wenig" : "wenig"
}
def attributive(adjective, gender=MALE, role=SUBJECT, article=None):
""" For a predicative adjective, returns the attributive form (lowercase).
In German, the attributive is formed with -e, -em, -en, -er or -es,
depending on gender (masculine, feminine, neuter or plural) and role
(nominative, accusative, dative, genitive).
"""
w, g, c, a = \
adjective.lower(), gender[:1].lower(), role[:3].lower(), article and article.lower() or None
if w in adjective_attributive:
return adjective_attributive[w]
if a is None \
or a in ("mir", "dir", "ihm") \
or a in ("ein", "etwas", "mehr") \
or a.startswith(("all", "mehrer", "wenig", "viel")):
return w + adjectives_strong.get((g, c), "")
if a.startswith(("ein", "kein")) \
or a.startswith(("mein", "dein", "sein", "ihr", "Ihr", "unser", "euer")):
return w + adjectives_mixed.get((g, c), "")
if a in ("arm", "alt", "all", "der", "die", "das", "den", "dem", "des") \
or a.startswith((
"derselb", "derjenig", "jed", "jeglich", "jen", "manch",
"dies", "solch", "welch")):
return w + adjectives_weak.get((g, c), "")
# Default to strong inflection.
return w + adjectives_strong.get((g, c), "")
def predicative(adjective):
""" Returns the predicative adjective (lowercase).
In German, the attributive form preceding a noun is always used:
"ein kleiner Junge" => strong, masculine, nominative,
"eine schöne Frau" => mixed, feminine, nominative,
"der kleine Prinz" => weak, masculine, nominative, etc.
The predicative is useful for lemmatization.
"""
w = adjective.lower()
if len(w) > 3:
for suffix in ("em", "en", "er", "es", "e"):
if w.endswith(suffix):
b = w[:max(-len(suffix), -(len(w) - 3))]
if b.endswith("bl"): # plausibles => plausibel
b = b[:-1] + "el"
if b.endswith("pr"): # propres => proper
b = b[:-1] + "er"
return b
return w
#### COMPARATIVE & SUPERLATIVE #####################################################################
COMPARATIVE = "er"
SUPERLATIVE = "st"
def grade(adjective, suffix=COMPARATIVE):
""" Returns the comparative or superlative form of the given (inflected) adjective.
"""
b = predicative(adjective)
# groß => großt, schön => schönst
if suffix == SUPERLATIVE and b.endswith(("s", "ß")):
suffix = suffix[1:]
# große => großere, schönes => schöneres
return adjective[:len(b)] + suffix + adjective[len(b):]
def comparative(adjective):
return grade(adjective, COMPARATIVE)
def superlative(adjective):
return grade(adjective, SUPERLATIVE)
#print(comparative("schönes"))
#print(superlative("schönes"))
#print(superlative("große"))