Как можно упростить часть речевых тегов, возвращаемых французским POS-тегером Стэнфорда? Довольно просто прочитать английское предложение в NLTK, найти часть речи каждого слова, а затем использовать map_tag () для упрощения набора тегов:

#!/usr/bin/python
# -*- coding: utf-8 -*-

import os
from nltk.tag.stanford import POSTagger
from nltk.tokenize import word_tokenize
from nltk.tag import map_tag

#set java_home path from within script. Run os.getenv("JAVA_HOME") to test java_home
os.environ["JAVA_HOME"] = "C:\\Program Files\\Java\\jdk1.7.0_25\\bin"

english = u"the whole earth swarms with living beings, every plant, every grain and leaf, supports the life of thousands."

path_to_english_model = "C:\\Text\\Professional\\Digital Humanities\\Packages and Tools\\Stanford Packages\\stanford-postagger-full-2014-08-27\\stanford-postagger-full-2014-08-27\\models\\english-bidirectional-distsim.tagger"
path_to_jar = "C:\\Text\\Professional\\Digital Humanities\\Packages and Tools\\Stanford Packages\\stanford-postagger-full-2014-08-27\\stanford-postagger-full-2014-08-27\\stanford-postagger.jar"

#define english and french taggers
english_tagger = POSTagger(path_to_english_model, path_to_jar, encoding="utf-8")

#each tuple in list_of_english_pos_tuples = (word, pos)
list_of_english_pos_tuples = english_tagger.tag(word_tokenize(english))

simplified_pos_tags_english = [(word, map_tag('en-ptb', 'universal', tag)) for word, tag in list_of_english_pos_tuples]

print simplified_pos_tags_english

#output = [(u'the', u'DET'), (u'whole', u'ADJ'), (u'earth', u'NOUN'), (u'swarms', u'NOUN'), (u'with', u'ADP'), (u'living', u'NOUN'), (u'beings', u'NOUN'), (u',', u'.'), (u'every', u'DET'), (u'plant', u'NOUN'), (u',', u'.'), (u'every', u'DET'), (u'grain', u'NOUN'), (u'and', u'CONJ'), (u'leaf', u'NOUN'), (u',', u'.'), (u'supports', u'VERB'), (u'the', u'DET'), (u'life', u'NOUN'), (u'of', u'ADP'), (u'thousands', u'NOUN'), (u'.', u'.')]

Но я не уверен, как сопоставить французские теги, возвращаемые следующим кодом, с универсальным набором тегов:

#!/usr/bin/python
# -*- coding: utf-8 -*-

import os
from nltk.tag.stanford import POSTagger
from nltk.tokenize import word_tokenize
from nltk.tag import map_tag

#set java_home path from within script. Run os.getenv("JAVA_HOME") to test java_home
os.environ["JAVA_HOME"] = "C:\\Program Files\\Java\\jdk1.7.0_25\\bin"

french = u"Chaque plante, chaque graine, chaque particule de matière organique contient des milliers d'atomes animés."

path_to_french_model = "C:\\Text\\Professional\\Digital Humanities\\Packages and Tools\\Stanford Packages\\stanford-postagger-full-2014-08-27\\stanford-postagger-full-2014-08-27\\models\\french.tagger"
path_to_jar = "C:\\Text\\Professional\\Digital Humanities\\Packages and Tools\\Stanford Packages\\stanford-postagger-full-2014-08-27\\stanford-postagger-full-2014-08-27\\stanford-postagger.jar"

french_tagger = POSTagger(path_to_french_model, path_to_jar, encoding="utf-8")

list_of_french_pos_tuples = french_tagger.tag(word_tokenize(french))

#up to this point all is well, but I'm not sure how to successfully create a simplified pos tagset with the French tuples
simplified_pos_tags_french = [(word, map_tag('SOME_ARGUMENT', 'universal', tag)) for word, tag in list_of_french_pos_tuples]
print simplified_pos_tags_french

Кто-нибудь знает, как упростить набор тегов по умолчанию, используемый французской моделью в Stanford POS tagger? Я был бы признателен за любые идеи, которые другие могут предложить по этому вопросу.