X-Git-Url: http://iramuteq.org/git?a=blobdiff_plain;ds=sidebyside;f=corpusNG.py;h=fa13a8b916b3762c806c9096fe0ff21722c76337;hb=b1bfe11bdde3e2499b2246f1841499a64480cabb;hp=a950d0f2ba23f6c2a687d8e1bbbb9af4569f41a8;hpb=81594f689f1e191599d96a2a503fbb5529df69d7;p=iramuteq
diff --git a/corpusNG.py b/corpusNG.py
index a950d0f..fa13a8b 100644
--- a/corpusNG.py
+++ b/corpusNG.py
@@ -3,6 +3,9 @@
import codecs
import os
+import gettext
+_ = gettext.gettext
+import locale
import sys
from time import time
from functions import decoupercharact, ReadDicoAsDico, DoConf
@@ -16,6 +19,7 @@ from uuid import uuid4
from chemins import PathOut
from dialog import CorpusPref
from functions import ReadLexique, ReadDicoAsDico
+from colors import colors
import datetime
@@ -123,6 +127,19 @@ class Corpus :
res = self.cformes.execute('SELECT uces FROM uces where id=? ORDER BY id;', (`wordid`,))
return list(itertools.chain(*[[int(val) for val in row[0].split()] if not isinstance(row[0], int) else [row[0]] for row in res]))
+ def getformeuceseff(self, formeid) :
+ if isinstance(formeid, basestring) :
+ formeid = self.formes[formeid].ident
+ res = self.cformes.execute('SELECT uces FROM uces where id=? ORDER BY id;', (`formeid`,))
+ uces = list(itertools.chain(*[[int(val) for val in row[0].split()] if not isinstance(row[0], int) else [row[0]] for row in res]))
+ query = 'SELECT eff FROM eff where id=%i ORDER BY id' % formeid
+ res = self.cformes.execute(query)
+ eff = list(itertools.chain(*[[int(val) for val in row[0].split()] if not isinstance(row[0], int) else [row[0]] for row in res]))
+ formeuceeff = {}
+ for i, uce in enumerate(uces) :
+ formeuceeff[uce] = formeuceeff.get(uce, 0) + eff[i]
+ return formeuceeff
+
def getlemuces(self, lem) :
formesid = ', '.join([`val` for val in self.lems[lem].formes])
query = 'SELECT uces FROM uces where id IN (%s) ORDER BY id' % formesid
@@ -133,7 +150,7 @@ class Corpus :
uces = self.getlemuces(lem)
return list(set([self.getucefromid(val).uci for val in uces]))
- def getlemuceseff(self, lem) :
+ def getlemuceseff(self, lem, luces = None) :
formesid = ', '.join([`val` for val in self.lems[lem].formes])
query = 'SELECT uces FROM uces where id IN (%s) ORDER BY id' % formesid
res = self.cformes.execute(query)
@@ -146,6 +163,9 @@ class Corpus :
lemuceeff[uce] = lemuceeff.get(uce, 0) + eff[i]
return lemuceeff
+ def getlemclustereff(self, lem, cluster) :
+ return len(list(set(self.lc[cluster]).intersection(self.getlemuces(lem))))
+
def getlemeff(self, lem) :
return self.lems[lem].freq
@@ -247,7 +267,7 @@ class Corpus :
self.iduces = dict([[uce.ident, uce] for uci in self.ucis for uce in uci.uces])
def make_lexitable(self, mineff, etoiles) :
- tokeep = [lem for lem in self.lems if self.lems[lem].freq > mineff]
+ tokeep = [lem for lem in self.lems if self.lems[lem].freq >= mineff]
etuces = [[] for et in etoiles]
for uci in self.ucis :
get = list(set(uci.etoiles).intersection(etoiles))
@@ -381,6 +401,28 @@ class Corpus :
actpara = self.iduces[uce[0]].para
ident += 1
f.write('\n'.join([self.ucis[self.iduces[uce[0]].uci].paras[ident].encode(self.parametres['syscoding']), uce[1].encode(self.parametres['syscoding'])]) + '\n')
+
+ def export_corpus_classes(self, outf, alc = True, lem = False) :
+ ucecl = {}
+ for i, lc in enumerate(self.lc) :
+ for uce in lc :
+ ucecl[uce] = i + 1
+ for uce in self.lc0 :
+ ucecl[uce] = 0
+ res = self.getalluces()
+ self.make_iduces()
+ with open(outf, 'w') as f :
+ for uce in res :
+ guce = uce[1]
+ actuci = self.iduces[uce[0]].uci
+ if lem :
+ guce = ' '.join([self.formes[forme].lem for forme in guce.split()])
+ if alc :
+ etline = ' '.join(self.ucis[self.iduces[uce[0]].uci].etoiles + ['*classe_%i' % ucecl[uce[0]]])
+ else :
+ etline = ' '.join(['<' + '='.join(et.split('_')) + '>' for et in self.ucis[self.iduces[uce[0]].uci].etoiles[1:]])
+ f.write(etline.encode(self.parametres['syscoding']) + '\n')
+ f.write(guce.encode(self.parametres['syscoding']) + '\n\n')
def make_and_write_sparse_matrix_from_uces(self, actives, outfile, listuce = False) :
log.info('make_and_write_sparse_matrix_from_uces %s' % outfile)
@@ -447,7 +489,9 @@ class Corpus :
def parse_active(self, gramact, gramsup = None) :
log.info('parse actives')
for lem in self.lems :
- if self.lems[lem].gram in gramact :
+ if lem.startswith('_') and lem.endswith('_') :
+ self.lems[lem].act = 2
+ elif self.lems[lem].gram in gramact :
self.lems[lem].act = 1
elif gramsup is not None :
if self.lems[lem].gram in gramsup :
@@ -457,10 +501,10 @@ class Corpus :
else :
self.lems[lem].act = 2
- def make_actives_limit(self, limit) :
+ def make_actives_limit(self, limit, key = 1) :
if self.idformes is None :
self.make_idformes()
- return [lem for lem in self.lems if self.getlemeff(lem) >= limit and self.lems[lem].act == 1]
+ return [lem for lem in self.lems if self.getlemeff(lem) >= limit and self.lems[lem].act == key]
def make_actives_nb(self, nbmax, key) :
log.info('make_actives_nb : %i - %i' % (nbmax,key))
@@ -485,7 +529,8 @@ class Corpus :
lim -= 1
else :
stop = nbmax - 1
- log.info('nb actives = %i - eff min = %i ' % (stop, lim))
+ lim = effs[stop]
+ log.info('nb actives = %i - eff min = %i ' % (stop + 1, lim))
return [val[1] for val in allactives[0:stop + 1]], lim
def make_and_write_profile(self, actives, ucecl, fileout) :
@@ -508,18 +553,30 @@ class Corpus :
et = etoile.split('_')
if et[0] in det :
try :
- if et[1] in det[et[0]] :
- det[et[0]][et[1]] += 1
+ endet = '_'.join(et[1:])
+ if endet in det[et[0]] :
+ det[et[0]][endet] += 1
else :
- det[et[0]][et[1]] = 1
+ det[et[0]][endet] = 1
except IndexError :
det[et[0]] += 1
else :
try :
- det[et[0]] = {et[1] :1}
+ endet = '_'.join(et[1:])
+ det[et[0]] = {endet :1}
except IndexError :
det[et[0]] = 1
- print det
+ return det
+
+ def make_etline(self, listet) :
+ etuces = [[] for et in listet]
+ for uci in self.ucis :
+ get = list(set(uci.etoiles).intersection(listet))
+ if len(get) > 1 :
+ return '2 variables sur la meme ligne'
+ elif get != [] :
+ etuces[listet.index(get[0])] += [uce.ident for uce in uci.uces]
+ return etuces
def make_and_write_profile_et(self, ucecl, fileout) :
@@ -528,6 +585,31 @@ class Corpus :
with open(fileout, 'w') as f :
f.write('\n'.join([';'.join([etoile] + [`len(set(self.getucesfrometoile(etoile)).intersection(classe))` for classe in ucecl]) for etoile in etoiles]).encode(self.parametres['syscoding']))
+ def make_colored_corpus(self) :
+ ucecl = {}
+ for i, lc in enumerate(self.lc) :
+ for uce in lc :
+ ucecl[uce] = i + 1
+ for uce in self.lc0 :
+ ucecl[uce] = 0
+ color = ['black'] + colors[len(self.lc) - 1]
+ txt = '''
+
+
+''' % sys.getdefaultencoding()
+ res = self.getalluces()
+ self.make_iduces()
+ actuci = ''
+ actpara = False
+ for uce in res :
+ if self.iduces[uce[0]].uci != actuci :
+ actuci = self.iduces[uce[0]].uci
+ txt += '
' + ' '.join(self.ucis[self.iduces[uce[0]].uci].etoiles) + '
'
+ txt += '' % (color[ucecl[uce[0]]]) + uce[1] + '
'
+ else :
+ txt += '' % (color[ucecl[uce[0]]]) + uce[1] + '
'
+ return txt + '\n'
+
def count_from_list(self, l, d) :
for val in l :
if val in d :
@@ -536,6 +618,15 @@ class Corpus :
d[val] = 1
return d
+ def count_from_list_cl(self, l, d, a, clnb) :
+ for val in l :
+ if val in d :
+ d[val][a] += 1
+ else :
+ d[val] = [0] * clnb
+ d[val][a] = 1
+ return d
+
def find_segments(self, taille_segment, taille_limite) :
d = {}
for uce in self.getalluces() :
@@ -560,9 +651,35 @@ class Corpus :
l = l[-taille_limite:]
return l
+ def make_segments_profile(self, fileout, lenmin = 3, lenmax = 10, effmin = 50, lem = False) :
+ d = {}
+ for b, classe in enumerate(self.lc) :
+ for uce in self.getconcorde(classe) :
+ uce = uce[1].split()
+ if lem :
+ uce = [self.formes[forme].lem for forme in uce]
+ for taille_segment in range(lenmin,lenmax) :
+ d =self.count_from_list_cl([' '.join(uce[i:i+taille_segment]) for i in range(len(uce)-(taille_segment - 1))], d, b, len(self.lc))
+ result = [[seg] + [str(val) for val in d[seg]] for seg in d if sum(d[seg]) >= effmin]
+ with open(fileout, 'w') as f :
+ f.write('\n'.join([';'.join(line) for line in result]))
+
+ def make_proftype(self, outf) :
+ res = {}
+ for lem in self.lems :
+ gram = self.lems[lem].gram
+ if not gram in res :
+ res[gram] = [0 for val in self.lc]
+ lemuceeff = self.getlemuceseff(lem)
+ for i, classe in enumerate(self.lc) :
+ concern = set(classe).intersection(lemuceeff.keys())
+ res[gram][i] += sum([lemuceeff[uce] for uce in concern])
+ res = [[gram] + [`val` for val in res[gram]] for gram in res]
+ res.sort()
+ with open(outf, 'w') as f :
+ f.write('\n'.join([';'.join(line) for line in res]).encode(self.parametres['syscoding']))
-
def make_ucecl_from_R(self, filein) :
with open(filein, 'rU') as f :
c = f.readlines()
@@ -577,6 +694,28 @@ class Corpus :
self.lc = [[uce[0] for uce in self.lc if uce[1] == i] for i in range(clnb+1)]
self.lc0 = self.lc.pop(0)
#return ucecl
+
+ def get_stat_by_cluster(self, outf) :
+ log.info('get_stat_by_cluster')
+ t1 = time()
+ occurrences = dict([[i + 1, 0] for i in range(len(self.lc))])
+ formescl = dict([[i + 1, 0] for i in range(len(self.lc))])
+ hapaxcl = dict([[i + 1, 0] for i in range(len(self.lc))])
+ lenclasses = dict([[i+1,len(cl)] for i, cl in enumerate(self.lc)])
+ sets = [set(cl) for cl in self.lc]
+ for forme in self.formes :
+ formeuceeff = self.getformeuceseff(forme)
+ for i, classe in enumerate(self.lc) :
+ concern = sets[i].intersection(formeuceeff.keys())
+ if len(concern) :
+ occurrences[i+1] += sum([formeuceeff[uce] for uce in concern])
+ formescl[i+1] += 1
+ if self.formes[forme].freq == 1 :
+ hapaxcl[i+1] += 1
+ toprint = '\n'.join([';'.join([`i`, `occurrences[i]`, `formescl[i]`, `hapaxcl[i]`, `lenclasses[i]`, `float(hapaxcl[i])/float(formescl[i])`]) for i in occurrences])
+ with open(outf, 'w') as f :
+ f.write(toprint)
+ log.info('%f' % (time() - t1))
def gethapaxbyet(self, etoiles) :
hapaxuces = [self.getlemuces(forme)[0] for forme in self.lems if self.lems[forme].freq == 1]
@@ -689,7 +828,7 @@ def decouperlist(chaine, longueur, longueurOptimale) :
Si on trouve un '$', c'est fini.
Sinon, on cherche le meilleur candidat. C'est-Ã -dire le rapport poids/distance le plus important.
"""
- separateurs = [[u'.', 6.0], [u'?', 6.0], [u'!', 6.0], [u'£', 6.0], [u':', 5.0], [u';', 4.0], [u',', 1.0], [u' ', 0.01]]
+ separateurs = [[u'.', 6.0], [u'?', 6.0], [u'!', 6.0], [u'£$£', 6.0], [u':', 5.0], [u';', 4.0], [u',', 1.0], [u' ', 0.01]]
dsep = dict([[val[0],val[1]] for val in separateurs])
trouve = False # si on a trouvé un bon séparateur
iDecoupe = 0 # indice du caractere ou il faut decouper
@@ -771,7 +910,7 @@ class BuildCorpus :
if self.corpus.parametres['keep_ponct'] :
self.ponctuation_espace = [' ', '']
else :
- self.ponctuation_espace = [' ','.', u'£', ';', '?', '!', ',', ':','']
+ self.ponctuation_espace = [' ','.', u'£$£', ';', '?', '!', ',', ':','']
self.cleans = []
self.tolist = self.corpus.parametres.get('tolist', 0)
self.buildcleans()
@@ -794,13 +933,18 @@ class BuildCorpus :
def dobuild(self) :
t1 = time()
- self.read_corpus(self.infile)
- self.indexdb()
- self.corpus.parametres['ira'] = self.corpus.pathout['Corpus.cira']
- self.time = time() - t1
- self.dofinish()
- DoConf().makeoptions(['corpus'],[self.corpus.parametres], self.corpus.pathout['Corpus.cira'])
- log.info('time : %f' % (time() - t1))
+ try :
+ self.read_corpus(self.infile)
+ except Warning, args :
+ log.info('pas kool %s' % args)
+ raise Warning
+ else :
+ self.indexdb()
+ self.corpus.parametres['ira'] = self.corpus.pathout['Corpus.cira']
+ self.time = time() - t1
+ self.dofinish()
+ DoConf().makeoptions(['corpus'],[self.corpus.parametres], self.corpus.pathout['Corpus.cira'])
+ log.info('time : %f' % (time() - t1))
def connect(self) :
self.conn_f = sqlite3.connect(self.corpus.pathout['formes.db'])
@@ -854,8 +998,9 @@ class BuildCorpus :
self.cleans.append(self.dolower)
if self.corpus.parametres.get('firstclean', 1) :
self.cleans.append(self.firstclean)
- self.rule = self.corpus.parametres.get('keep_caract', u"^a-zA-Z0-9à ÃâÃäÃáÃéÃèÃêÃëÃìÃîÃïÃòÃôÃöÃùÃûÃüÃçÃÃÅÅâñ.:,;!?*'_-")
- self.cleans.append(self.docharact)
+ if self.corpus.parametres['charact'] :
+ self.rule = self.corpus.parametres.get('keep_caract', u"^a-zA-Z0-9à ÃâÃäÃáÃéÃèÃêÃëÃìÃîÃïÃòÃôÃöÃùÃûÃüÃçÃÃÅÅâñ.:,;!?*'_")
+ self.cleans.append(self.docharact)
if self.corpus.parametres.get('expressions', 1) :
self.cleans.append(self.make_expression)
if self.corpus.parametres.get('apos', 1) :
@@ -886,7 +1031,7 @@ class BuildCorpus :
def firstclean(self, txt) :
txt = txt.replace(u'â',"'")
txt = txt.replace(u'Å', u'oe')
- return txt.replace('...',u' £ ').replace('?',' ? ').replace('.',' . ').replace('!', ' ! ').replace(',',' , ').replace(';', ' ; ').replace(':',' : ')
+ return txt.replace('...',u' £$£ ').replace('?',' ? ').replace('.',' . ').replace('!', ' ! ').replace(',',' , ').replace(';', ' ; ').replace(':',' : ').replace(u'â¦', ' £$£ ')
def make_cleans(self, txt) :
for clean in self.cleans :
@@ -930,11 +1075,9 @@ class BuildCorpus :
class BuildFromAlceste(BuildCorpus) :
- #def __init___(self, infile, parametres_corpus) :
- # BuildCorpus.__init__(self, infile, parametres_corpus)
-
-
def read_corpus(self, infile) :
+ if self.dlg is not None :
+ self.dlg.Pulse('textes : 0 - segments : 0')
self.limitshow = 0
self.count = 1
if self.corpus.parametres['ucimark'] == 0 :
@@ -945,28 +1088,50 @@ class BuildFromAlceste(BuildCorpus) :
iduci = -1
idpara = -1
iduce = -1
- with codecs.open(infile, 'rU', self.encoding) as f :
- for line in f :
- if self.testuci(line) :
- iduci += 1
- if txt != [] :
- iduce, idpara = self.treattxt(txt, iduce, idpara, iduci - 1)
- txt = []
- self.corpus.ucis.append(Uci(iduci, line))
- else :
- self.corpus.ucis.append(Uci(iduci, line))
- elif line.startswith(u'-*') :
- if txt != [] :
- iduce, idpara = self.treattxt(txt, iduce, idpara, iduci)
- txt = []
- idpara += 1
- self.corpus.ucis[-1].paras.append(line.split()[0])
- elif line.strip() != '' and iduci != -1 :
- txt.append(line)
- if txt != [] :
- iduce, idpara = self.treattxt(txt, iduce, idpara, iduci)
- del(txt)
- self.backup_uce()
+ try :
+ with codecs.open(infile, 'r', self.encoding) as f :
+ for linenb, line in enumerate(f) :
+ line = line.rstrip('\n\r')
+ if self.testuci(line) :
+ iduci += 1
+ if txt != [] :
+ iduce, idpara = self.treattxt(txt, iduce, idpara, iduci - 1)
+ txt = []
+ self.corpus.ucis.append(Uci(iduci, line))
+ else :
+ if iduci > 0 :
+ if self.corpus.ucis[-1].uces == [] :
+ log.info(u'Empty text : %i' % linenb)
+ iduci -= 1
+ self.corpus.ucis.pop()
+ #raise Exception("EmptyText %i" % linenb)
+ self.corpus.ucis.append(Uci(iduci, line))
+ if self.dlg is not None :
+ if not (iduci + 1) % 10 :
+ self.dlg.Pulse('textes : %i - segments : %i' % (iduci + 1, iduce +1))
+ elif line.startswith(u'-*') :
+ if iduci != -1 :
+ if txt != [] :
+ iduce, idpara = self.treattxt(txt, iduce, idpara, iduci)
+ txt = []
+ idpara += 1
+ self.corpus.ucis[-1].paras.append(line.split()[0])
+ else :
+ raise Exception('paragrapheOT')
+ elif line.strip() != '' and iduci != -1 :
+ txt.append(line)
+ if txt != [] and iduci != -1 :
+ iduce, idpara = self.treattxt(txt, iduce, idpara, iduci)
+ del(txt)
+ else :
+ raise Exception("EmptyText")
+ if iduci != -1 and iduce != -1:
+ self.backup_uce()
+ else :
+ log.info(_(u"No Text in corpora. Are you sure of the formatting ?"))
+ raise Exception('TextBeforeTextMark')
+ except UnicodeDecodeError :
+ raise Exception("CorpusEncoding")
def treattxt(self, txt, iduce, idpara, iduci) :
if self.corpus.parametres.get('ucemethod', 0) == 2 and self.corpus.parametres['douce']:
@@ -991,13 +1156,6 @@ class BuildFromAlceste(BuildCorpus) :
for word in uce :
self.last += 1
self.corpus.add_word(word)
- if self.dlg is not None :
- if self.limitshow > self.count :
- self.dlg.Pulse('uci: %i - uce : %i' % (iduci + 1, iduce +1))
- self.count += 1
- self.limitshow = 0
- else :
- self.limitshow = self.last / 100000
log.debug(' '.join([`iduci`,`idpara`,`iduce`]))
if self.last > self.lim :
self.backup_uce()
@@ -1009,27 +1167,13 @@ class BuildFromAlceste(BuildCorpus) :
if douce :
out = []
reste, texte_uce, suite = self.decouper(self.prep_txt(txt), self.ucesize + 15, self.ucesize)
-# print 'reste'
-# print reste
-# print 'texte_uce'
-# print texte_uce
-# print 'suite'
-# print suite
while reste :
uce = ' '.join([val for val in texte_uce if val not in self.ponctuation_espace])
if uce != '' :
out.append(uce)
reste, texte_uce, suite = self.decouper(suite, self.ucesize + 15, self.ucesize)
-# print 'reste'
-# print reste
-# print 'texte_uce'
-# print texte_uce
-# print 'suite'
-# print suite
-
uce = ' '.join([val for val in texte_uce if val not in self.ponctuation_espace])
if uce != '' :
- print 'RESTEE UUCEEEEEEEEEEEEE', uce
out.append(uce)
return out
else :
@@ -1058,6 +1202,9 @@ class Builder :
ReadLexique(self.parent, lang = parametres['lang'])
self.parent.expressions = ReadDicoAsDico(self.parent.DictPath.get(parametres['lang']+'_exp', 'french_exp'))
self.parametres = parametres
+ else :
+ if self.dlg is not None :
+ self.dlg.Destroy()
dial.Destroy()
def doanalyse(self) :