def get_meter(self, meter=None):
if not meter:
if self.meter:
meter = self.meter
elif hasattr(self, '_Text__bestparses') and self.__bestparses:
return self.get_meter(sorted(self.__bestparses.keys())[0])
else:
import Meter
meter = Meter.genDefault()
#print '>> no meter specified. defaulting to this meter:'
#print meter
elif type(meter) in [str, unicode]:
meter = self.config['meters'][meter]
else:
pass
return meter
def get_meter(self,meter=None): if not meter: if self.meter: meter=self.meter elif hasattr(self,'_Text__bestparses') and self.__bestparses: return self.get_meter(sorted(self.__bestparses.keys())[0]) else: import Meter meter=Meter.genDefault() #print '>> no meter specified. defaulting to this meter:' #print meter elif type(meter) in [str,unicode]: meter= self.config['meters'][meter] else: pass return meter
def report(self,meter=None,include_bounded=False,reverse=True):
""" Print all parses and their violations in a structured format. """
ReportStr = ''
if not meter:
from Meter import Meter
meter=Meter.genDefault()
if (hasattr(self,'allParses')):
self.om(unicode(self))
allparses=self.allParses(meter=meter,include_bounded=include_bounded)
numallparses=len(allparses)
#allparses = reversed(allparses) if reverse else allparses
for pi,parseList in enumerate(allparses):
line=self.iparse2line(pi).txt
#parseList.sort(key = lambda P: P.score())
hdr="\n\n"+'='*30+'\n[line #'+str(pi+1)+' of '+str(numallparses)+']: '+line+'\n\n\t'
ftr='='*30+'\n'
ReportStr+=self.om(hdr+meter.printParses(parseList,reverse=reverse).replace('\n','\n\t')[:-1]+ftr,conscious=False)
else:
for child in self.children:
if type(child)==type([]): continue
ReportStr+=child.report()
return ReportStr
def assess(fn,meter=None,key_meterscheme=None, key_line='line',key_parse='parse'):
#from prosodic import Text
import prosodic as p
Text=p.Text
if not meter:
import Meter
meter=Meter.genDefault()
p.config['print_to_screen']=0
def parse2list(parse):
l=[]
for i,x in enumerate(parse):
if not l or l[-1]!=x:
l+=[x]
else:
l[-1]+=x
return l
def get_num_sylls_correct(parse_human,parse_comp):
maxlen=max([len(parse_comp),len(parse_human)])
#parse_human=parse2list(parse_human)
#parse_comp=parse2list(parse_comp)
#parse_comp_forzip = parse_comp + ['x' for x in range(maxlen-len(parse_comp))]
#parse_human_forzip = parse_human + ['x' for x in range(maxlen-len(parse_human))]
parse_comp_forzip = parse_comp + ''.join(['x' for x in range(maxlen-len(parse_comp))])
parse_human_forzip = parse_human + ''.join(['x' for x in range(maxlen-len(parse_human))])
## sylls correct?
_sylls_iscorrect=[]
#print '\t'.join(parse_human_forzip)
#print '\t'.join(parse_comp_forzip)
for syll1,syll2 in zip(parse_human_forzip,parse_comp_forzip):
syll_iscorrect = int(syll1==syll2)
_sylls_iscorrect+=[syll_iscorrect]
return _sylls_iscorrect
import codecs
ld=read_ld(fn)
fn_split = fn.split('.')
ofn_split=fn_split[:-1] + ['evaluated','meter='+meter.id] + [fn_split[-1]]
ofn_split_ot=fn_split[:-1] + ['evaluated', 'ot','meter='+meter.id] + [fn_split[-1]]
ofn='.'.join(ofn_split)
ofn_ot='.'.join(ofn_split_ot)
def _print(dx):
print
for k,v in sorted(dx.items()):
print k,'\t',v
print 'HUMAN :','\t'.join(dx['parse_human'])
print 'PROSODIC:','\t'.join(dx['parse_comp'])
print ' ','\t'.join(['*' if x!=y else ' ' for x,y in zip(dx['parse_human'],dx['parse_comp'])])
print
def _recapitalize(parse,code):
code=' '.join([x for x in code])
parse=parse.replace('|',' ').replace('.',' ')
newparse=[]
for s,c in zip(parse.split(),code.split()):
if c=='w':
newparse+=[s.lower()]
else:
newparse+=[s.upper()]
return ' '.join(newparse)
def _writegen():
lines_iscorrect=[]
lines_iscorrect_control=[]
lines_iscorrect_control2=[]
lines_iscorrect_human2=[]
sylls_iscorrect_control=[]
sylls_iscorrect_control2=[]
sylls_iscorrect_human2=[]
sylls_iscorrect=[]
lines_iscorrect_nonbounded=[]
otf=open(ofn_ot,'w')
otf_nl=0
for di,d in enumerate(ld):
line=d[key_line]
parse_human=''.join([x for x in d[key_parse].lower() if x in ['s','w']])
if not parse_human: continue
t=Text(line)
t.parse(meter=meter)
#if not t.isParsed: continue
parse_comp=t.parse_str(viols=False, text=False).replace('|','')
#if len(parse_comp) != len(parse_human): continue
parse_str=t.parse_str(viols=False, text=True)
parses_comp = [x.replace('|','') for x in t.parse_strs(viols=False,text=False)]
parse_human2=''.join([x for x in d.get('parse_human2','').lower() if x in ['s','w']])
#parse_human,parse_human2=parse_human2,parse_human
### OT
if not otf_nl:
header=['','','']
for c in meter.constraints: header+=['[*'+c.name+']']
otf.write('\t'.join(header)+'\n')
humans = [parse_human]
if parse_human2: humans+=[parse_human2]
for _i,_parses in enumerate(t.allParses()):
if not _parses: continue
_parses.sort(key=lambda _P: (-humans.count(_P.str_meter()), _P.totalCount))
if not humans.count(_parses[0].str_meter()):
# is the good parse in the bounded ones?
for _bndp in t.boundParses()[_i]:
if _bndp.str_meter() in humans:
_parses.insert(0,_bndp)
for _pi,_parse in enumerate(_parses):
otf_nl+=1
code=_parse.str_meter()
row=[line.encode('utf-8',errors='ignore') if not _pi else '', str(_parse) + (' [*Bounded]' if _parse.isBounded else ''), str(humans.count(code)) if code in humans else '']
for c in meter.constraints: row+=[str(_parse.constraintCounts[c]) if _parse.constraintCounts[c] else '']
otf.write('\t'.join(row)+'\n')
parse_comp_dummy2 = ''.join(['w' if not i%2 else 's' for i in range(len(parse_comp))])
if key_meterscheme:
if d[key_meterscheme]=='iambic':
parse_comp_dummy = ('ws'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='trochaic':
parse_comp_dummy = ('sw'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='anapestic':
parse_comp_dummy = ('wws'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='dactylic':
parse_comp_dummy = ('sww'*100)[:len(parse_comp)]
else:
parse_comp_dummy=parse_comp_dummy2
else:
parse_comp_dummy=parse_comp_dummy2
## sylls correct?
this_sylls_correct = get_num_sylls_correct(parse_human, parse_comp)
this_sylls_correct_dummy = get_num_sylls_correct(parse_human, parse_comp_dummy)
this_sylls_correct_dummy2 = get_num_sylls_correct(parse_human, parse_comp_dummy2)
if parse_human2: this_sylls_correct_human2 = get_num_sylls_correct(parse_human, parse_human2)
num_sylls_correct=sum(this_sylls_correct)
num_sylls_correct_dummy = sum(this_sylls_correct_dummy)
num_sylls_correct_dummy2 = sum(this_sylls_correct_dummy2)
if parse_human2: num_sylls_correct_human2 = sum(this_sylls_correct_human2)
sylls_iscorrect+=this_sylls_correct
sylls_iscorrect_control+=this_sylls_correct_dummy
sylls_iscorrect_control2+=this_sylls_correct_dummy2
if parse_human2: sylls_iscorrect_human2+=this_sylls_correct_human2
# line correct?
line_iscorrect=int(parse_comp == parse_human)
lines_iscorrect+=[line_iscorrect]
line_iscorrect_dummy = int(parse_comp_dummy == parse_human)
line_iscorrect_dummy2 = int(parse_comp_dummy2 == parse_human)
if parse_human2: line_iscorrect_human2 = int(parse_human2 == parse_human)
lines_iscorrect_control+=[line_iscorrect_dummy]
lines_iscorrect_control2+=[line_iscorrect_dummy2]
if parse_human2: lines_iscorrect_human2+=[line_iscorrect_human2]
# line at least in list of nonbounded parses?
line_iscorrect_nonbounded=int(parse_human in parses_comp)
lines_iscorrect_nonbounded+=[line_iscorrect_nonbounded]
parse_stress = []
for w in t.words():
for x in w.stress:
parse_stress += ['w' if x=='U' else 's']
parse_stress=''.join(parse_stress)
odx=d
odx['parse_human']=parse_human
if parse_human2: odx['parse_human2']=parse_human2
odx['parse_comp']=parse_comp
odx['parses_comp_nonbounded']=' | '.join(parses_comp)
odx['num_sylls']=len(parse_human)
odx['num_sylls_correct']=num_sylls_correct
odx['num_sylls_correct_control']=num_sylls_correct_dummy
odx['num_sylls_correct_control_iambic']=num_sylls_correct_dummy2
if parse_human2:
odx['num_sylls_correct_human2']=num_sylls_correct_human2
odx['perc_sylls_correct_human2']=num_sylls_correct_human2 / float(len(parse_human))
odx['line_iscorrect_human2']=line_iscorrect_human2
odx['perc_sylls_correct']=num_sylls_correct / float(len(parse_human))
odx['perc_sylls_correct_control']=num_sylls_correct_dummy / float(len(parse_human))
odx['perc_sylls_correct_control_iambic']=num_sylls_correct_dummy2 / float(len(parse_human))
odx['line_iscorrect']=line_iscorrect
odx['line_iscorrect_dummy']=line_iscorrect_dummy
odx['line_iscorrect_dummy_iambic']=line_iscorrect_dummy2
odx['line_is_in_nonbounded_parses']=line_iscorrect_nonbounded
odx['parse_str_human']=_recapitalize(parse_str, parse_human)
odx['parse_str_compu']=_recapitalize(parse_str, parse_comp)
odx['parse_str_stress']=_recapitalize(parse_str, parse_stress)
odx['prosody_ipa']=' '.join([w.str_ipasyllstress() for w in t.words()])
odx['prosody_stress']=' '.join([w.stress for w in t.words()])
odx['meter_info']=str(t.meter).replace('\n',' ').replace('\t',' ')
sumconstr=0
for k,v in t.constraintViolations(use_weights=False,normalize=False).items():
odx['constraint_'+k]=v
sumconstr+=v
odx['constraint_SUM_VIOL']=sumconstr
#if not line_iscorrect and line_iscorrect_dummy:
#if len(parse_comp) != len(parse_human):
#if len(parse_human)>len(parse_comp):
_print(odx)
yield odx
print
print '##'*10
print 'RESULTS SUMMARY'
print '##'*10
perc_sylls_correct = sum(sylls_iscorrect) / float(len(sylls_iscorrect)) * 100
perc_lines_correct = sum(lines_iscorrect) / float(len(lines_iscorrect)) * 100
perc_lines_correct_control = sum(lines_iscorrect_control) / float(len(lines_iscorrect_control)) * 100
perc_sylls_correct_control = sum(sylls_iscorrect_control) / float(len(sylls_iscorrect_control)) * 100
perc_lines_correct_nonbound = sum(lines_iscorrect_nonbounded) / float(len(lines_iscorrect_nonbounded)) * 100
print 'PERCENT SYLLABLES CORRECT:',round(perc_sylls_correct,2),'% [vs.',round(perc_sylls_correct_control,2),'% for control]'
print 'PERCENT LINES CORRECT:',round(perc_lines_correct,2),'% [vs.',round(perc_lines_correct_control,2),'% for control]'
print 'PERCENT LINES IN AVAILABLE NONBOUNDED PARSES:',round(perc_lines_correct_nonbound,2),'%'
writegen(ofn, _writegen)
def assess(fn,meter=None,key_meterscheme=None, key_line='line',key_parse='parse'):
#from prosodic import Text
import prosodic as p
Text=p.Text
if not meter:
import Meter
meter=Meter.genDefault()
p.config['print_to_screen']=0
def parse2list(parse):
l=[]
for i,x in enumerate(parse):
if not l or l[-1]!=x:
l+=[x]
else:
l[-1]+=x
return l
def get_num_sylls_correct(parse_human,parse_comp):
maxlen=max([len(parse_comp),len(parse_human)])
#parse_human=parse2list(parse_human)
#parse_comp=parse2list(parse_comp)
#parse_comp_forzip = parse_comp + ['x' for x in range(maxlen-len(parse_comp))]
#parse_human_forzip = parse_human + ['x' for x in range(maxlen-len(parse_human))]
parse_comp_forzip = parse_comp + ''.join(['x' for x in range(maxlen-len(parse_comp))])
parse_human_forzip = parse_human + ''.join(['x' for x in range(maxlen-len(parse_human))])
## sylls correct?
_sylls_iscorrect=[]
#print '\t'.join(parse_human_forzip)
#print '\t'.join(parse_comp_forzip)
for syll1,syll2 in zip(parse_human_forzip,parse_comp_forzip):
syll_iscorrect = int(syll1==syll2)
_sylls_iscorrect+=[syll_iscorrect]
return _sylls_iscorrect
import codecs
ld=read_ld(fn)
fn_split = fn.split('.')
ofn_split=fn_split[:-1] + ['evaluated','meter='+meter.id] + [fn_split[-1]]
ofn_split_ot=fn_split[:-1] + ['evaluated', 'ot','meter='+meter.id] + [fn_split[-1]]
ofn='.'.join(ofn_split)
ofn_ot='.'.join(ofn_split_ot)
def _print(dx):
print()
for k,v in sorted(dx.items()):
print(k,'\t',v)
print('HUMAN :','\t'.join(dx['parse_human']))
print('PROSODIC:','\t'.join(dx['parse_comp']))
print(' ','\t'.join(['*' if x!=y else ' ' for x,y in zip(dx['parse_human'],dx['parse_comp'])]))
print()
def _recapitalize(parse,code):
code=' '.join([x for x in code])
parse=parse.replace('|',' ').replace('.',' ')
newparse=[]
for s,c in zip(parse.split(),code.split()):
if c=='w':
newparse+=[s.lower()]
else:
newparse+=[s.upper()]
return ' '.join(newparse)
def _writegen():
lines_iscorrect=[]
lines_iscorrect_control=[]
lines_iscorrect_control2=[]
lines_iscorrect_human2=[]
sylls_iscorrect_control=[]
sylls_iscorrect_control2=[]
sylls_iscorrect_human2=[]
sylls_iscorrect=[]
lines_iscorrect_nonbounded=[]
otf=open(ofn_ot,'w')
otf_nl=0
for di,d in enumerate(ld):
line=d[key_line]
parse_human=''.join([x for x in d[key_parse].lower() if x in ['s','w']])
if not parse_human: continue
t=Text(line)
t.parse(meter=meter)
#if not t.isParsed: continue
parse_comp=t.parse_str(viols=False, text=False).replace('|','')
#if len(parse_comp) != len(parse_human): continue
parse_str=t.parse_str(viols=False, text=True)
parses_comp = [x.replace('|','') for x in t.parse_strs(viols=False,text=False)]
parse_human2=''.join([x for x in d.get('parse_human2','').lower() if x in ['s','w']])
#parse_human,parse_human2=parse_human2,parse_human
### OT
if not otf_nl:
header=['','','']
for c in meter.constraints: header+=['[*'+c.name+']']
otf.write('\t'.join(header)+'\n')
humans = [parse_human]
if parse_human2: humans+=[parse_human2]
for _i,_parses in enumerate(t.allParses()):
if not _parses: continue
_parses.sort(key=lambda _P: (-humans.count(_P.str_meter()), _P.totalCount))
if not humans.count(_parses[0].str_meter()):
# is the good parse in the bounded ones?
for _bndp in t.boundParses()[_i]:
if _bndp.str_meter() in humans:
_parses.insert(0,_bndp)
for _pi,_parse in enumerate(_parses):
otf_nl+=1
code=_parse.str_meter()
row=[line.encode('utf-8',errors='ignore') if not _pi else '', str(_parse) + (' [*Bounded]' if _parse.isBounded else ''), str(humans.count(code)) if code in humans else '']
for c in meter.constraints: row+=[str(_parse.constraintCounts[c]) if _parse.constraintCounts[c] else '']
otf.write('\t'.join(row)+'\n')
parse_comp_dummy2 = ''.join(['w' if not i%2 else 's' for i in range(len(parse_comp))])
if key_meterscheme:
if d[key_meterscheme]=='iambic':
parse_comp_dummy = ('ws'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='trochaic':
parse_comp_dummy = ('sw'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='anapestic':
parse_comp_dummy = ('wws'*100)[:len(parse_comp)]
elif d[key_meterscheme]=='dactylic':
parse_comp_dummy = ('sww'*100)[:len(parse_comp)]
else:
parse_comp_dummy=parse_comp_dummy2
else:
parse_comp_dummy=parse_comp_dummy2
## sylls correct?
this_sylls_correct = get_num_sylls_correct(parse_human, parse_comp)
this_sylls_correct_dummy = get_num_sylls_correct(parse_human, parse_comp_dummy)
this_sylls_correct_dummy2 = get_num_sylls_correct(parse_human, parse_comp_dummy2)
if parse_human2: this_sylls_correct_human2 = get_num_sylls_correct(parse_human, parse_human2)
num_sylls_correct=sum(this_sylls_correct)
num_sylls_correct_dummy = sum(this_sylls_correct_dummy)
num_sylls_correct_dummy2 = sum(this_sylls_correct_dummy2)
if parse_human2: num_sylls_correct_human2 = sum(this_sylls_correct_human2)
sylls_iscorrect+=this_sylls_correct
sylls_iscorrect_control+=this_sylls_correct_dummy
sylls_iscorrect_control2+=this_sylls_correct_dummy2
if parse_human2: sylls_iscorrect_human2+=this_sylls_correct_human2
# line correct?
line_iscorrect=int(parse_comp == parse_human)
lines_iscorrect+=[line_iscorrect]
line_iscorrect_dummy = int(parse_comp_dummy == parse_human)
line_iscorrect_dummy2 = int(parse_comp_dummy2 == parse_human)
if parse_human2: line_iscorrect_human2 = int(parse_human2 == parse_human)
lines_iscorrect_control+=[line_iscorrect_dummy]
lines_iscorrect_control2+=[line_iscorrect_dummy2]
if parse_human2: lines_iscorrect_human2+=[line_iscorrect_human2]
# line at least in list of nonbounded parses?
line_iscorrect_nonbounded=int(parse_human in parses_comp)
lines_iscorrect_nonbounded+=[line_iscorrect_nonbounded]
parse_stress = []
for w in t.words():
for x in w.stress:
parse_stress += ['w' if x=='U' else 's']
parse_stress=''.join(parse_stress)
odx=d
odx['parse_human']=parse_human
if parse_human2: odx['parse_human2']=parse_human2
odx['parse_comp']=parse_comp
odx['parses_comp_nonbounded']=' | '.join(parses_comp)
odx['num_sylls']=len(parse_human)
odx['num_sylls_correct']=num_sylls_correct
odx['num_sylls_correct_control']=num_sylls_correct_dummy
odx['num_sylls_correct_control_iambic']=num_sylls_correct_dummy2
if parse_human2:
odx['num_sylls_correct_human2']=num_sylls_correct_human2
odx['perc_sylls_correct_human2']=num_sylls_correct_human2 / float(len(parse_human))
odx['line_iscorrect_human2']=line_iscorrect_human2
odx['perc_sylls_correct']=num_sylls_correct / float(len(parse_human))
odx['perc_sylls_correct_control']=num_sylls_correct_dummy / float(len(parse_human))
odx['perc_sylls_correct_control_iambic']=num_sylls_correct_dummy2 / float(len(parse_human))
odx['line_iscorrect']=line_iscorrect
odx['line_iscorrect_dummy']=line_iscorrect_dummy
odx['line_iscorrect_dummy_iambic']=line_iscorrect_dummy2
odx['line_is_in_nonbounded_parses']=line_iscorrect_nonbounded
odx['parse_str_human']=_recapitalize(parse_str, parse_human)
odx['parse_str_compu']=_recapitalize(parse_str, parse_comp)
odx['parse_str_stress']=_recapitalize(parse_str, parse_stress)
odx['prosody_ipa']=' '.join([w.str_ipasyllstress() for w in t.words()])
odx['prosody_stress']=' '.join([w.stress for w in t.words()])
odx['meter_info']=str(t.meter).replace('\n',' ').replace('\t',' ')
sumconstr=0
for k,v in list(t.constraintViolations(use_weights=False,normalize=False).items()):
odx['constraint_'+k]=v
sumconstr+=v
odx['constraint_SUM_VIOL']=sumconstr
#if not line_iscorrect and line_iscorrect_dummy:
#if len(parse_comp) != len(parse_human):
#if len(parse_human)>len(parse_comp):
_print(odx)
yield odx
print()
print('##'*10)
print('RESULTS SUMMARY')
print('##'*10)
perc_sylls_correct = sum(sylls_iscorrect) / float(len(sylls_iscorrect)) * 100
perc_lines_correct = sum(lines_iscorrect) / float(len(lines_iscorrect)) * 100
perc_lines_correct_control = sum(lines_iscorrect_control) / float(len(lines_iscorrect_control)) * 100
perc_sylls_correct_control = sum(sylls_iscorrect_control) / float(len(sylls_iscorrect_control)) * 100
perc_lines_correct_nonbound = sum(lines_iscorrect_nonbounded) / float(len(lines_iscorrect_nonbounded)) * 100
print('PERCENT SYLLABLES CORRECT:',round(perc_sylls_correct,2),'% [vs.',round(perc_sylls_correct_control,2),'% for control]')
print('PERCENT LINES CORRECT:',round(perc_lines_correct,2),'% [vs.',round(perc_lines_correct_control,2),'% for control]')
print('PERCENT LINES IN AVAILABLE NONBOUNDED PARSES:',round(perc_lines_correct_nonbound,2),'%')
writegen(ofn, _writegen)
