def is_how(const_tree):
#returns is_how, obj, how
cpy = const_tree.copy(deep=True)
badadv = ['all', 'almost', 'already', 'also', 'basically', 'further', 'finally', 'generally', 'greatly', 'however', 'initially', 'just', 'later', 'largely', 'longer', 'mostly', 'meanwhile', 'often', 'only', 'perhaps', 'now', 'then', 'typically']
#try by
vbphr = getVP(const_tree, len(const_tree))
(thing, bydoing) = findBy(const_tree, vbphr)
if thing != None:
return thing, " ".join(bydoing.leaves())
else:
#try adv
(thing, how) = findAdv(cpy, getVP(cpy, len(cpy)))
if thing == None or leftmost(how)[0].lower() in badadv:
return None, None
else:
return cpy, " ".join(how.leaves())
def who(const_tree, nertags):
res = is_who(const_tree, nertags)
q = ''
if res:
vbphr = getVP(const_tree, len(const_tree))
if vbphr:
vp = ''
for v in vbphr:
vp += ' ' + ' '.join(v.leaves())
q = "Who" + vp + "?"
return q
return None
def reason_cause(const_tree):
#something, cause
for phr in const_tree: #if by is before subject
if phr.label() == 'SBAR' and leftmost(phr)[0] == 'because':
const_tree.remove(phr)
return const_tree, phr
#if by is within VP
vp = getVP(const_tree, len(const_tree))
if vp:
for phr in vp:
if phr.label() == 'SBAR' and leftmost(phr)[0] == 'because':
reason = phr
vp.remove(reason)
return const_tree, " ".join(reason.leaves())
return None, None
def is_who(const_tree, nertags):
cpy = const_tree.copy(deep=True)
vbphr = getVP(const_tree, len(const_tree))
nphr = getNP(const_tree, len(const_tree))
if not (vbphr and nphr):
return None
poss_who = ''
for n in nphr:
poss_who += ' ' + ' '.join(n.leaves())
spacy_nlp = spacy.load('en')
noun = spacy_nlp(poss_who)
who = None
for c in noun.ents:
if (c.label_ == ("PERSON" or "ORG" or "GPE" or "NORP")):
who = c
if (who == None):
return None
return str(who)
def is_time(const_tree, nertags, answering):
timetags = ['DATE', 'TIME']
vp = getVP(const_tree, len(const_tree))
for phr in const_tree: #if on same level
if phr.label() == 'PP':
npTree = searchPhrase(phr, 'NP')
if len(npTree) > 0:
np = ' '.join(npTree[0].leaves()).split()
else:
break
for n in nertags:
if n[1] in timetags and np[0] == (n[0]):
const_tree.remove(phr)
return const_tree, " ".join(phr.leaves())
if vp:
for phr in vp:
#TODO should be more robust but works for now
npTree = searchPhrase(phr, 'NP')
if phr.label() == 'PP':
if len(npTree) > 0:
np = ' '.join(npTree[0].leaves()).split()
else:
break
for n in nertags:
if n[1] in timetags and np[0] == (n[0]):
vp.remove(phr)
return const_tree, " ".join(phr.leaves())
if answering:
res = ''
for n in nertags:
if n[1] in timetags:
res += n[0] + ' '
if len(res) > 0 and n[1] not in timetags:
break
if len(res) > 0:
return const_tree, res
return None, None
def is_howmany(const_tree, nertags):
numtags = ["NUMBER", "CARDINAL"]
np = getNP(const_tree, len(const_tree))
vp = getVP(const_tree, len(const_tree))
const_tree1 = const_tree.copy(deep=True)
(const_tree2, preps) = searchAndRem(const_tree1, 'PP')
preps_lst = []
if preps:
preps_lst = ' '.join(preps[0].leaves()).split()
if vp:
subnpTree = searchPhrase(vp, 'NP')
verb = ' '.join(' '.join(vp[0].leaves()).split())
lem_verb = lem(u'' + verb, u'VERB')[0]
if len(subnpTree) > 0:
subnp = ' '.join(subnpTree[0].leaves()).split()
for n in nertags:
if (n[1] in numtags and subnp[0] == (n[0])):
restnp = (' '.join(subnp)).replace(subnp[0], "")
doform = getDoForm(vp[0])[0].lower() + getDoForm(vp[0])[1:]
q_body = ' '.join(' '.join(np.leaves()).split())
q_body_lower = q_body[0].lower() + q_body[1:]
(const_tree3, subpreps) = searchAndRem(vp, 'PP')
if subpreps:
p_str = ' '.join(' '.join(
subpreps[0].leaves()).split())
p_str_lower = p_str[0].lower() + p_str[1:]
ques = 'How many' + restnp + ' ' + doform + ' ' + q_body_lower + ' ' + lem_verb + ' ' + p_str_lower + '?'
else:
ques = 'How many' + restnp + ' ' + doform + ' ' + q_body_lower + ' ' + lem_verb + '?'
return ques, subnp[0]
if (len(preps_lst) > 0):
for n in nertags:
if (n[1] in numtags and n[0] in preps_lst):
np_string = ' '.join(' '.join(np.leaves()).split())
np_string_lower = np_string[0].lower() + np_string[1:]
vp_string = ' '.join(' '.join(vp.leaves()).split())
no_prep = vp_string.replace(' '.join(preps_lst), "")
rest_preps = ' '.join(preps_lst).replace(n[0], "").replace(
preps_lst[0], "")
no_prep_verb = no_prep.replace(verb, "")
try:
(prep_tree, det) = searchAndRem(preps, 'DT')
rest_preps2 = ' '.join(' '.join(
prep_tree[0].leaves()).split())
final_preps = rest_preps2.replace(
' '.join(' '.join(det[0].leaves()).split()),
"").replace(n[0], "").replace(preps_lst[0], "")
ques = 'How many' + final_preps + ' ' + verb + ' ' + np_string_lower + no_prep_verb + " " + preps_lst[
0] + "?"
except:
ques = 'How many' + rest_preps + ' ' + verb + ' ' + np_string_lower + no_prep_verb + " " + preps_lst[
0] + "?"
return ques, n[0]
np_list = ' '.join(np.leaves()).split()
for n in nertags:
if (n[1] in numtags and n[0] in np_list):
q_body = ' '.join(' '.join(const_tree.leaves()).split())
q_body1 = q_body.replace(n[0], "")
q_body1_lower = q_body1[0].lower() + q_body1[1:]
q_body2 = q_body1_lower
try:
(np_tree, det) = searchAndRem(np, 'DT')
rest_np = ' '.join(' '.join(np_tree.leaves()).split())
vp_str = ' '.join(' '.join(vp.leaves()).split())
q_body2 = rest_np.replace(
' '.join(' '.join(det[0].leaves()).split()),
"").replace(n[0], "") + ' ' + vp_str
except:
pass
ques = 'How many ' + q_body2 + "?"
return ques, n[0]
return None, None
return None, None