def registry_infer_exts():
extensions.register_parts(
'*',
{
'advmod': lambda c, t: (4, "extract_for('word', 'advmod')"),
'det': lambda c, t: (4, "extract_for('plain', 'det')"),
'obl': lambda c, t: (4, "extract_for('chunk', 'verb:obl')"),
'cop': lambda c, t: (2, "cop='c_aux'"),
'head_amod': lambda c, t: (2, "head_amod=interr('what')"),
# $ spt 'A cobra fala com o menino.'
'nsubj': lambda c, t: induce_subj(c, t),
# obj如果有kindof匹配结果, 下面的注册项不会被执行
# 'obj': lambda c, t: induce_measure(c, t),
'nummod': lambda c, t:
(4, f"extract_for('plain+number', '{c.name}')"),
'obl:arg': lambda c, t: induce_propn(c, t),
})
def get_verb_interr(c: DomainToken, part: Text):
from sagas.nlu.inspectors_dataset import get_interrogative
from sagas.nlu.transliterations import translits
word = translits.translit(c.text.split('/')[0], 'ko')
rep = get_interrogative(word, 'ko')
if rep:
return 4, f"interr_root('{rep}')"
else:
return 4, "interr_root('??')"
extensions.register_parts(
'ko', {
'nsubj':
lambda c, t: [(4, "extract_for('plain', 'nsubj')"),
(2, "nsubj=agency")],
'obl':
lambda c, t: get_nouns_spec(c, 'obl'),
})
extensions.register_domains(
'ko',
{
# Testcases:
# $ sko '우리는 피자와 스파게티가 필요해요.' -> spec
# $ sko '이번 주말에 벌써 계획이 있어요?' -> interr_root
'verb': lambda c, t: get_verb_spec(c, t) or get_verb_interr(c, t),
})
class Rules_ko(LangSpecBase):
@staticmethod
def get_from_to(c: InferPart, t):
results = []
if 'カラ' in c.domain.rels:
results.append((4, "extract_for('plain', 'カラ')"))
if 'マデ' in c.domain.rels:
results.append((4, "extract_for('plain', 'マデ')"))
return results
extensions.register_parts(
'ja', {
'時間': lambda c, t:
(4, "extract_for('plain+date_search+date_parse', '時間')"),
'ガ': lambda c, t: (4, "extract_for('plain', 'ガ')"),
'デ': lambda c, t: (4, "extract_for('plain', 'デ')"),
'ニ': lambda c, t: (4, "extract_for('plain', 'ニ')"),
'修飾': lambda c, t: (4, "extract_for('plain+number', '修飾')"),
'カラ': lambda c, t: get_from_to(c, t),
'マデ': lambda c, t: get_from_to(c, t),
})
def get_verb_interr(c: DomainToken, part: Text):
return 4, "interr_root('??')"
extensions.register_domains('ja', {
'verb': lambda c, t: get_verb_interr(c, t),
})
registry_named_exprs(
import logging
logger = logging.getLogger(__name__)
pron = ['c_pron', 'c_det']
def induce_pron(c: InferPart, t: Text):
if c.lemma == '-PRON-':
return [(4, f"extract_for('plain', '{c.name}')"),
(2, f"{c.name}=pron")]
extensions.register_parts(
'lt', {
'nsubj': lambda c, t: induce_pron(c, t),
'nmod': lambda c, t: induce_pron(c, t),
})
class Rules_lt(LangSpecBase):
# def opts(self):
# return {'engine':'spacy'}
@staticmethod
def prepare(meta: Dict[Text, Any]):
tc.emp('yellow', '.. Rules_lt(Lithuanian, 立陶宛语) prepare phrase')
def verb_rules(self):
pat, actions_obj = (self.pat, self.actions_obj)
import logging
from sagas.nlu.tool_base import LangToolBase
logger = logging.getLogger(__name__)
def induce_dim(c:InferPart, t:Text, dim:Text):
pat=c.domain.pattern(t)
r = pat(**{c.name:dateins(dim)})
logger.debug(f"t:{t}, dim:{dim}, result:{r[1]}, {r[0]}")
if r[1]:
return 2, f"{c.name}=dateins('{dim}')"
extensions.register_parts('pt',{
# $ spt 'Eu preciso disso até amanhã.'
'advmod': lambda c,t: induce_dim(c, t, 'time'),
# $ spt 'Ele está entre meu irmão e minha irmã.'
'case': lambda c,t: predict_pos(c, t, 'c_adp'),
})
class Rules_pt(LangToolBase):
@staticmethod
def prepare(meta: Dict[Text, Any]):
tc.emp('yellow', '.. Rules_pt(Portuguese, 葡萄牙语) prepare phrase')
def verb_rules(self):
pat, actions_obj = (self.pat, self.actions_obj)
self.collect(pats=[
# $ se 'I want to watch a movie'
pat(5, name='behave_willing_ev').verb(behaveof('want', 'v'),
pred_any_path('xcomp/obj', 'social_event', 'n')),
from sagas.nlu.tool_base import LangToolBase
logger = logging.getLogger(__name__)
def head_interr(c:InferPart, part:Text):
from sagas.nlu.inspectors_dataset import get_interrogative
rep=get_interrogative(c.lemma, 'ru')
if rep:
return 2, f"{part}=interr('{rep}')"
else:
return 4, f"extract_for('plain', '{part}')"
extensions.register_parts('ru',{
'nsubj': lambda c,t: [(4, "extract_for('plain', 'nsubj')"),
(2, "nsubj=agency")],
'head_csubj': lambda c,t: head_interr(c, 'head_csubj'),
})
class Rules_ru(LangToolBase):
@staticmethod
def prepare(meta: Dict[Text, Any]):
tc.emp('yellow', '.. Rules_ru(Russian, 俄语) prepare phrase')
def verb_rules(self):
pat, actions_obj = (self.pat, self.actions_obj)
self.collect(pats=[
# $ sru 'У Вас есть сигареты?'
# $ sru 'У Вас есть пепельница?'
pat(5, name='behave_exist').verb(extract_for('plain', 'nsubj'),
behaveof('exist', 'v'), nsubj=agency),