def __init__(self, lang='en'):
from sagas.nlu.corenlp_helper import get_nlp
self.lang = lang
# spacy_model=lang_spacy_mappings[lang][0]
# self.spacy_nlp=spacy.load(spacy_model)
self.core_nlp = get_nlp(lang)
def handle_digest():
"""
$ curl -XPOST -H 'Content-Type: application/json' -d '{"sents":"Die Eltern mögen den Käse."}' http://localhost:8090/digest
will get: {"root": "mögen", "verbs": [["mögen", "mögen"]]}
:return:
"""
from sagas.nlu.corenlp_helper import get_nlp
# print ("request is json?", request.is_json)
content = request.get_json()
sents = content['sents']
lang = content['lang']
print(lang, sents)
# dumped = json_to_string({"ok": 'JSON posted'})
nlp = get_nlp(lang)
doc = nlp(sents)
root, root_idx = get_doc_root_and_idx(doc)
# subj(nsubj), obj(iobj, dobj, pobj)
# pobj : object of a preposition，介词的宾语
# obl类似pobj, obl关系用于名义（名词，代词，名词短语），作为非核心（倾斜）参数或附件。
# 这意味着它在功能上对应于附加在动词，形容词或其他副词上的状语。
rs = get_root_rel(doc, root_idx, ['subj', 'obj', 'cop', 'obl'])
data = {'lang': lang, 'root': root, 'verbs': get_doc_verbs(doc)}
for el in rs:
data[el[0]] = el[1]
# data_y=yaml.dump(data, default_flow_style=True,Dumper=KludgeDumper,encoding=None)
data_y = json.dumps(data, ensure_ascii=False)
return data_y
def words_table(sents, lang):
from sagas.nlu.corenlp_helper import get_nlp
nlp = get_nlp(lang)
doc = nlp(sents)
sentence = doc.sentences[0]
rows = [[
word.text, word.lemma, word.upos, word.xpos, word.dependency_relation,
word.governor, word.feats
] for word in sentence.words]
return sagas.to_df(
rows, ['text', 'lemma', 'upos', 'xpos', 'dep', 'head', 'feats'])
def viz(sents, lang='fr'):
"""
viz("Tu as choisi laquelle tu vas manger ?")
viz('I am a student', 'en')
viz('彼らは3月に訪ねて来ます。', 'ja')
:param sents:
:param lang:
:return:
"""
nlp = get_nlp(lang)
doc = nlp(sents)
cv = EnhancedViz(shape='egg', size='8,5', fontsize=20)
sentence = doc.sentences[0]
return cv.analyse_doc(sentence, None)
def parse(self, sents, lang='en'):
"""
$ python -m sagas.nlu.corenlp_helper parse 'Barack Obama was born in Hawaii.' en
$ python -m sagas.nlu.corenlp_helper parse 'Θα το θέλατε με ρύζι;' el
$ parse 'Die weiße Fläche ist aus dem All sichtbar.' de
$ parse '私は高校生です。' ja
$ parse 'Yo tengo una casa en México.' es
:param sents:
:param lang:
:return:
"""
# routine=langs[lang]
analyse(sents, get_nlp(lang))
def verb_domains(self, sents, lang='en'):
"""
$ python -m sagas.nlu.uni_parser verb_domains "Barack Obama was born in Hawaii." en
# 我有一只阿比西尼亚猫
$ python -m sagas.nlu.uni_parser verb_domains "I have an Abyssinian cat." en
$ python -m sagas.nlu.uni_parser verb_domains 'Что ты обычно ешь на ужин?' ru
$ python -m sagas.nlu.uni_parser verb_domains 'Die Zeitschrift erscheint monatlich.' de
# 测试多个动词(过滤掉从句的动词):
$ python -m sagas.nlu.uni_parser verb_domains 'Tu as choisi laquelle tu vas manger ?' fr
:param sents:
:param lang:
:return:
"""
from sagas.nlu.corenlp_helper import get_nlp
serial_numbers = '❶❷❸❹❺❻❼❽❾❿'
nlp = get_nlp(lang)
doc = nlp(sents)
# 分析依赖关系, 自下而上, 可用于抽取指定关系的子节点集合, 比如此例中的'nsubj:pass'和'obl'
# word.governor即为当前word的parent
sent = doc.sentences[0]
rs = get_verb_domain(sent)
# r=rs[0]
for num, r in enumerate(rs):
# print(json.dumps(r, indent=2, ensure_ascii=False))
print(serial_numbers[num], '-' * 50)
# print(r['verb'], r['index'])
print(r['word'], r['index'])
# df=sagas.to_df(r[0]['domains'], ['rel', 'index', 'text', 'children'])
df = sagas.to_df(
r['domains'],
['rel', 'index', 'text', 'lemma', 'children', 'features'])
sagas.print_df(df)
for stem in r['stems']:
if stem[0] == 'obj':
print('object ->', ' '.join(stem[1]))
def __call__(self, sents):
from sagas.nlu.corenlp_helper import get_nlp
nlp = get_nlp(self.lang)
doc = nlp(sents)
return CoreNlpSentImpl(doc.sentences[0], text=sents)