def test_conf():
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
from sagas.nlu.anal_corpus import model_info
f = build_anal_tree('Ördek filin üzerinde.', 'tr', 'stanza')
# f.draw()
assert 'AnalNode_tr' == type(f).__name__
assert 'Doc' == type(f.doc).__name__
def test_zh():
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
f = build_anal_tree('附近有什么好吃的饭馆', 'zh', 'ltp')
f.draw()
assert (f / 'obj').deprel == 'obj'
assert ['饭馆', '饭馆'
] == [f.resolve_rel("obj").text,
f.resolve_rels("obj*")[0].text]
def test_and_or():
f = build_anal_tree('list some restaurants', 'en', 'stanza')
f.draw()
target = f.model().target
print(target.text, target.types)
assert target.match('InstitutePlace|场所')
assert target.match('location: reside|住下; location: eat|吃')
assert not target.match('+location: reside|住下; location: eat|吃')
def test_match_absent_part():
f = build_anal_tree('Nós estudamos.', 'pt', 'stanza')
f.draw()
r = match(f, behave_(_, _1 << 'study', _2 << _,
_), lambda arg, v1, v2: [v1.text, v2.text],
behave_(_, _1 << 'study', _, _), lambda arg, v1: [v1.text],
behave_(_, _1 << 'learn', _2 << _, _),
lambda arg, v1, v2: [v1.text, v2.text], _, None)
assert ['estudamos'] == r
def test_match_embed_expr():
f = build_anal_tree('We expect them to change their minds', 'en', 'stanza')
f.draw()
r = match(
f,
behave_(_, _1 << 'expect|期望', _, _,
rel_('xcomp') == behave_(_, _2 << 'change|变', 'mind')),
lambda arg, v1, v2: [v1.text, v2.text], _, None)
assert ['expect', 'change'] == r
def test_restaurants(sents, lang, engine):
f = build_anal_tree(sents, lang, engine)
f.draw()
target = f.model().target
agents = match_agents(target)
assert len(agents) == 1
rs = agents[0](f)
pprint(rs)
assert len(rs) > 1
def descrip(self, sents, lang, engine=None):
"""
$ python -m sagas.nlu.anal_corpus descrip 'Karpet di kantor saya abu-abu.' id
$ sid 'Celana ini bisa diperbesar.'
:param sents:
:param lang:
:param engine:
:return:
"""
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
from sagas.conf.conf import cf
f = build_anal_tree(sents, lang, cf.engine(lang))
f.draw()
model = f.model()
model_info(model)
def test_behave_matcher():
f = build_anal_tree('We expect them to change their minds', 'en', 'stanza')
f.draw()
model = f.rels('xcomp')[0].model()
model_info(model)
assert (f/'nsubj').match(_)
assert (f/'obj').match('pos:pron')
assert (f / 'xcomp').match(behave_(_, 'change|变', 'mind'))
# assert (f/'xcomp').match(behave_(_, 'change|变', 'mind') |
# behave_(_, 'change|变', 'mental|精神'))
assert f.match(behave_(_, 'expect|期望', _, _))
assert not f.match(behave_(_, 'love', _, _))
assert f.match(behave_(_, 'expect|期望', _, _,
rel_('xcomp')==behave_(_, 'change|变', 'mind')))
assert f==behave_(_, 'expect|期望', _, _)
def test_match_expr():
f = build_anal_tree('Note the output is a string', 'en', 'stanza')
f.draw()
r=match(f,
behave_(_, 'perception|感知', _, _), lambda arg: 'perception',
behave_(_, 'perception|感知', desc_('result|结果', _), _), lambda arg: arg.text,
_, None
)
assert 'perception'==r
r = match(f,
behave_(_, 'unknown', _, _), lambda arg: 'unknown',
# behave_(_, 'perception|感知', desc_('result|结果', _), _),
# lambda arg: [arg.behave.text],
behave_(_, 'perception|感知', _1<<desc_('result|结果', _), _),
lambda arg, v1: [arg.behave.text, v1.target.text],
_, None
)
assert ['Note', 'output'] == r
# assert ['Note'] == r
r = match(f/'ccomp'/'nsubj',
'pos:noun', lambda arg: arg.text,
_, None
)
assert 'output' == r
r = match(f / 'ccomp' / 'nsubj',
'pos:pron', lambda arg: arg.text,
_, None
)
assert r is None
r = match(f,
behave_(_, 'perception|感知', _1<<_, _), lambda arg, v1: v1.text,
_, None
)
assert 'string' == r
r = match(f,
behave_(_, _1<<'perception|感知', _2 << _, _), lambda arg, v1, v2: [v1.text, v2.text],
_, None
)
assert ['Note', 'string'] == r
def do(self, sents, lang='en'):
"""
$ python -m sagas.modules.life.agents do 'list some restaurants' en
:param sents:
:param lang:
:return:
"""
f = build_anal_tree(sents, lang, cf.engine(lang))
tc.emp('cyan', '❏', f.doc.sents)
f.draw()
target = f.model().target
tc.emp('red', '☞', target.text, target.types)
for m in global_meta_ls():
print('-' * 25)
succ=match_agent(target, m, verbose=True)
if succ:
r = m(f)
if r:
pprint(r)
else:
print('result -> _')
def test_match_expr():
f = build_anal_tree('Note the output is a string', 'en', 'stanza')
f.draw()
r = match(f, behave_(_, _1 << 'perception|感知', _2 << _, _),
lambda arg, v1, v2: [v1.text, v2.text], _, None)
assert ['Note', 'string'] == r
r = match(
f,
behave_(_, _1 << 'unknown', _2 << _, _),
lambda *args: 'unknown',
# behave_(_, 'perception|感知', _1 << desc_('result|结果', _), _),
# lambda arg, v1: [arg.behave.text, v1.target.text],
behave_(_, _1 << 'perception|感知', desc_('result|结果', _), _),
lambda arg, v1: [arg.behave.text, v1.text],
behave_(_, _1 << 'unknown2', _2 << _, _),
lambda *args: 'unknown2',
_,
None)
assert ['Note', 'Note'] == r
assert 0 == len(_1.reqs)
def digest_verb(sents, lang, engine='stanza'):
f = build_anal_tree(sents, lang, engine)
words = findall_by_attr(f, name='upos', value='VERB')
if words:
print(sents, len(words))
rs = []
for w in words:
rs.append(proc_verb_subs(w))
def proc_text(t):
if lang in ('ko'):
return get_contrast(t, lang)
return t
succ = any(rs)
cl = 'blue' if succ else 'white'
tc.emp(
cl,
RenderTree(f, style=AsciiStyle()).by_attr(
lambda n:
f"{n.dependency_relation}: {proc_text(n.text)} {n.upos}"))
return succ
return False
def test_term(sents, lang, engine):
f = build_anal_tree(sents, lang, engine)
f.draw()
assert f.is_term('ref')
pprint(f.links[0].as_json())
assert f.links[0].name == 'ProductType'
def test_ner_rasa():
# run: simple/rasa-serv.sh
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
f = build_anal_tree('i would like to find an expensive restaurant', 'en',
'stanza')
assert f.get_by_index(7).ner.value == 'hi'
def test_ja():
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
f = build_anal_tree('コーヒーを ミルク付きで お願い します 。', 'ja', 'knp')
f.draw()
assert ['珈琲', '珈琲'] == [f.resolve_rel("obj").lemma,
f.resolve_rels("obj*")[0].lemma]
def test_as_desc():
from sagas.nlu.anal import build_anal_tree, Doc, AnalNode
f = build_anal_tree('Nuestro horario es de nueve a cinco.', 'es', 'stanza')
desc = f.as_desc()
assert (desc.subj.text, desc.subj_spec, desc.aux.lemma, desc.desc.as_num()) \
== ('horario', 'schedule', 'ser', 9)
def test_desc_matcher():
f = build_anal_tree('Note the output is a string', 'en', 'stanza')
f.draw()
assert f == behave_(_, 'perception|感知', _, _)
assert f == behave_(_, 'perception|感知', desc_('result|结果', _), _)
assert not f == behave_(_, 'perception|感知', desc_('food', _), _)
def test_term(sents, lang, engine):
f = build_anal_tree(sents, lang, engine)
f.draw()
# target = f.model().target
assert (f / 'nsubj').is_term('typ')