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.

284 lines
9.0 KiB
Python

5 years ago
#### PATTERN | FR ##################################################################################
# -*- coding: utf-8 -*-
# Copyright (c) 2013 University of Antwerp, Belgium
# Copyright (c) 2013 St. Lucas University College of Art & Design, Antwerp.
# Author: Tom De Smedt <tom@organisms.be>
# License: BSD (see LICENSE.txt for details).
# http://www.clips.ua.ac.be/pages/pattern
####################################################################################################
# French linguistical tools using fast regular expressions.
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
try:
MODULE = os.path.dirname(os.path.realpath(__file__))
except:
MODULE = ""
sys.path.insert(0, os.path.join(MODULE, "..", "..", "..", ".."))
# Import parser base classes.
from pattern.text import (
Lexicon, Model, Morphology, Context, Parser as _Parser, ngrams, pprint, commandline,
PUNCTUATION
)
# Import parser universal tagset.
from pattern.text import (
penntreebank2universal as _penntreebank2universal,
PTB, PENN, UNIVERSAL,
NOUN, VERB, ADJ, ADV, PRON, DET, PREP, ADP, NUM, CONJ, INTJ, PRT, PUNC, X
)
# Import parse tree base classes.
from pattern.text.tree import (
Tree, Text, Sentence, Slice, Chunk, PNPChunk, Chink, Word, table,
SLASH, WORD, POS, CHUNK, PNP, REL, ANCHOR, LEMMA, AND, OR
)
# Import sentiment analysis base classes.
from pattern.text import (
Sentiment as _Sentiment,
NOUN, VERB, ADJECTIVE, ADVERB,
MOOD, IRONY
)
# Import spelling base class.
from pattern.text import (
Spelling
)
# Import verb tenses.
from pattern.text import (
INFINITIVE, PRESENT, PAST, FUTURE,
FIRST, SECOND, THIRD,
SINGULAR, PLURAL, SG, PL,
INDICATIVE, IMPERATIVE, SUBJUNCTIVE, CONDITIONAL,
IMPERFECTIVE, PERFECTIVE, PROGRESSIVE,
IMPERFECT, PRETERITE,
PARTICIPLE, GERUND
)
# Import inflection functions.
from pattern.text.fr.inflect import (
pluralize, singularize, NOUN, VERB, ADJECTIVE,
verbs, conjugate, lemma, lexeme, tenses,
predicative, attributive
)
# Import all submodules.
from pattern.text.fr import inflect
sys.path.pop(0)
#--- FRENCH PARSER ---------------------------------------------------------------------------------
# The French parser is based on Lefff (Lexique des Formes Fléchies du Français).
# Benoît Sagot, Lionel Clément, Érice Villemonte de la Clergerie, Pierre Boullier.
# The Lefff 2 syntactic lexicon for French: architecture, acquisition.
# http://alpage.inria.fr/~sagot/lefff-en.html
# For words in Lefff that can have different part-of-speech tags,
# we used Lexique to find the most frequent POS-tag:
# http://www.lexique.org/
_subordinating_conjunctions = set((
"afin", "comme", "lorsque", "parce", "puisque", "quand", "que", "quoique", "si"
))
def penntreebank2universal(token, tag):
""" Converts a Penn Treebank II tag to a universal tag.
For example: comme/IN => comme/CONJ
"""
if tag == "IN" and token.lower() in _subordinating_conjunctions:
return CONJ
return _penntreebank2universal(token, tag)
ABBREVIATIONS = set((
"av.", "boul.", "C.-B.", "c.-à-d.", "ex.", "éd.", "fig.", "I.-P.-E.", "J.-C.",
"Ltee.", "Ltée.", "M.", "Me.", "Mlle.", "Mlles.", "MM.", "N.-B.", "N.-É.", "p.",
"S.B.E.", "Ste.", "T.-N.", "t.a.b."
))
# While contractions in English are optional,
# they are required in French:
replacements = {
"l'": "l' ", # le/la
"c'": "c' ", # ce
"d'": "d' ", # de
"j'": "j' ", # je
"m'": "m' ", # me
"n'": "n' ", # ne
"qu'": "qu' ", # que
"s'": "s' ", # se
"t'": "t' ", # te
"jusqu'": "jusqu' ",
"lorsqu'": "lorsqu' ",
"puisqu'": "puisqu' ",
# Same rule for Unicode apostrophe, see also Parser.find_tokens():
r"(l|c|d|j|m|n|qu|s|t|jusqu|lorsqu|puisqu)": "\\1&rsquo; "
}
replacements.update(((k.upper(), v.upper()) for k, v in list(replacements.items())))
def find_lemmata(tokens):
""" Annotates the tokens with lemmata for plural nouns and conjugated verbs,
where each token is a [word, part-of-speech] list.
"""
for token in tokens:
word, pos, lemma = token[0], token[1], token[0]
if pos.startswith(("DT", "PR", "WP")):
lemma = singularize(word, pos=pos)
if pos.startswith(("RB", "IN")) and (word.endswith(("'", "")) or word == "du"):
lemma = singularize(word, pos=pos)
if pos.startswith(("JJ",)):
lemma = predicative(word)
if pos == "NNS":
lemma = singularize(word)
if pos.startswith(("VB", "MD")):
lemma = conjugate(word, INFINITIVE) or word
token.append(lemma.lower())
return tokens
class Parser(_Parser):
def find_tokens(self, tokens, **kwargs):
kwargs.setdefault("abbreviations", ABBREVIATIONS)
kwargs.setdefault("replace", replacements)
s = _Parser.find_tokens(self, tokens, **kwargs)
s = [s.replace("&rsquo ;", "") if isinstance(s, str) else s for s in s]
return s
def find_lemmata(self, tokens, **kwargs):
return find_lemmata(tokens)
def find_tags(self, tokens, **kwargs):
if kwargs.get("tagset") in (PENN, None):
kwargs.setdefault("map", lambda token, tag: (token, tag))
if kwargs.get("tagset") == UNIVERSAL:
kwargs.setdefault("map", lambda token, tag: penntreebank2universal(token, tag))
return _Parser.find_tags(self, tokens, **kwargs)
class Sentiment(_Sentiment):
def load(self, path=None):
_Sentiment.load(self, path)
# Map "précaire" to "precaire" (without diacritics, +1% accuracy).
if not path:
for w, pos in list(dict.items(self)):
w0 = w
if not w.endswith(("à", "è", "é", "ê", "ï")):
w = w.replace("à", "a")
w = w.replace("é", "e")
w = w.replace("è", "e")
w = w.replace("ê", "e")
w = w.replace("ï", "i")
if w != w0:
for pos, (p, s, i) in pos.items():
self.annotate(w, pos, p, s, i)
parser = Parser(
lexicon = os.path.join(MODULE, "fr-lexicon.txt"),
frequency = os.path.join(MODULE, "fr-frequency.txt"),
morphology = os.path.join(MODULE, "fr-morphology.txt"),
context = os.path.join(MODULE, "fr-context.txt"),
default = ("NN", "NNP", "CD"),
language = "fr"
)
lexicon = parser.lexicon # Expose lexicon.
sentiment = Sentiment(
path = os.path.join(MODULE, "fr-sentiment.xml"),
synset = None,
negations = ("n'", "ne", "ni", "non", "pas", "rien", "sans", "aucun", "jamais"),
modifiers = ("RB",),
modifier = lambda w: w.endswith("ment"),
tokenizer = parser.find_tokens,
language = "fr"
)
spelling = Spelling(
path = os.path.join(MODULE, "fr-spelling.txt")
)
def tokenize(s, *args, **kwargs):
""" Returns a list of sentences, where punctuation marks have been split from words.
"""
return parser.find_tokens(s, *args, **kwargs)
def parse(s, *args, **kwargs):
""" Returns a tagged Unicode string.
"""
return parser.parse(s, *args, **kwargs)
def parsetree(s, *args, **kwargs):
""" Returns a parsed Text from the given string.
"""
return Text(parse(s, *args, **kwargs))
def tree(s, token=[WORD, POS, CHUNK, PNP, REL, LEMMA]):
""" Returns a parsed Text from the given parsed string.
"""
return Text(s, token)
def tag(s, tokenize=True, encoding="utf-8", **kwargs):
""" Returns a list of (token, tag)-tuples from the given string.
"""
tags = []
for sentence in parse(s, tokenize, True, False, False, False, encoding, **kwargs).split():
for token in sentence:
tags.append((token[0], token[1]))
return tags
def keywords(s, top=10, **kwargs):
""" Returns a sorted list of keywords in the given string.
"""
return parser.find_keywords(s, **dict({
"frequency": parser.frequency,
"top": top,
"pos": ("NN",),
"ignore": ("rt",)}, **kwargs))
def suggest(w):
""" Returns a list of (word, confidence)-tuples of spelling corrections.
"""
return spelling.suggest(w)
def polarity(s, **kwargs):
""" Returns the sentence polarity (positive/negative) between -1.0 and 1.0.
"""
return sentiment(s, **kwargs)[0]
def subjectivity(s, **kwargs):
""" Returns the sentence subjectivity (objective/subjective) between 0.0 and 1.0.
"""
return sentiment(s, **kwargs)[1]
def positive(s, threshold=0.1, **kwargs):
""" Returns True if the given sentence has a positive sentiment (polarity >= threshold).
"""
return polarity(s, **kwargs) >= threshold
split = tree # Backwards compatibility.
#---------------------------------------------------------------------------------------------------
# python -m pattern.fr xml -s "C'est l'exception qui confirme la règle." -OTCL
if __name__ == "__main__":
commandline(parse)