class FeatureExtractor(xml.sax.handler.ContentHandler):
def __init__(self, label, document, features, fd):
self.fd = fd
self.arff = Arff()
self.label = label
self.document = document
self.features = features
self.cur_label = None
self.cur_doc = None
self.cur_feat = []
self.docstack = 0
self.labstack = 0
self.featstack = 0
def startElement(self, name, attrs):
if name == self.label.split('[')[0]:
if self.cur_label != None:
self.labstack += 1
elif match(self.label, name, attrs):
self.labstack = 1
self.cur_label = attrs['n']
if name == self.document.split('[')[0]:
if self.cur_doc != None:
self.docstack += 1
elif match(self.document, name, attrs):
self.docstack = 1
self.cur_doc = {'label' : self.cur_label}
if name == 'f':
self.cur_feat.append(attrs.get('name', '?'))
if name == 'fs':
self.cur_feat.append(attrs.get('type', '?'))
if name == 'symbol' and self.cur_doc != None:
key = "%s=%s" % ("_".join(self.cur_feat), attrs['value'])
self.cur_doc[key.lstrip('?_')] = self.cur_doc.get('value', 0) + 1
def endElement(self, name):
if name == self.label.split('[')[0]:
if self.cur_label != None:
self.labstack -= 1
if self.labstack == 0 and self.cur_label != None:
self.cur_label = None
if name == self.document.split('[')[0]:
if self.cur_doc != None:
self.docstack -= 1
if self.docstack == 0 and self.cur_doc != None:
self.arff.add_datum(self.cur_doc)
self.cur_doc = None
if name in ['f', 'fs']:
self.cur_feat = self.cur_feat[0:-1]
## def characters(self, content):
## if self.write == True:
## self.fd.write("%s" % content.strip())
## elif len(content.strip()) > 0:
## self.fd.write(self.handler(content))
def endDocument(self):
self.arff.save(self.fd)
xml.sax.parse(options.input, FeatureExtractor(options.label, options.document, options.features, open(options.output, 'w')))
sys.exit()
a = Arff()
#doc = libxml2.parseFile(options.input)
doc = etree.parse(open(options.input))
doc.xinclude()
for labeled in doc.xpath(options.label):
label = labeled
if isinstance(labeled, etree._ElementStringResult):
labeled = labeled.getparent()
#labeled.xinclude()
for text in labeled.xpath(options.document):
if isinstance(text, etree._ElementStringResult):
text = text.getparent()
#print type(text)
#text.xincludeProcessTree()
attribs = {"LABEL" : label}
#print text
for feature in options.features:
seq = [str(x) for x in text.xpath(feature)]
#print seq
for i in set(seq):
attribs[i] = seq.count(i) / float(len(seq))
a.add_datum(attribs)
a.save(fd)