def _get_features(self, v="", v_corpus=None, cls2id=None, domain="src"):
_flist = []
_labellist_int = []
_labellist_str = []
_labelid = cls2id[v]
if v_corpus:
for sid, s in enumerate(v_corpus):
try:
fe = FeatureExtractor(s, verb=v)
if "chunk" in self.featuretypes:
fe.chunk()
if "3gram" in self.featuretypes:
fe.ngrams(n=3)
if "5gram" in self.featuretypes:
fe.ngrams(n=5)
if "7gram" in self.featuretypes:
fe.ngrams(n=7)
if "dep" in self.featuretypes:
fe.dependency()
if "srl" in self.featuretypes:
fe.srl()
if "ne" in self.featuretypes:
fe.ne()
if "errorprob" in self.featuretypes:
pass
if "topic" in self.featuretypes:
pass
augf = proc_easyadapt(fe.features, domain=domain)
_flist.append(augf)
_labellist_int.append(_labelid)
_labellist_str.append(v)
except ValueError:
logging.debug(pformat("CaseMaker feature extraction: couldn't find the verb"))
except:
print v
raise
else:
_flist.append(self.nullfeature)
_labellist_int.append(_labelid)
_labellist_str.append(v)
return _flist, _labellist_str, _labellist_int
def _get_features_tgt(self, v_corpus=None, cls2id=None, domain="tgt"):
_flist = []
_labellist_int = []
_labellist_str = []
for sid, sdic in enumerate(v_corpus):
v = sdic["label_corr"]
_labelid = cls2id[v]
try:
fe = SentenceFeatures(sdic["parsed_corr"], verb=v, v_idx=sdic["vidx_corr"])
if "chunk" in self.featuretypes:
fe.chunk()
if "3gram" in self.featuretypes:
fe.ngrams(n=3)
if "5gram" in self.featuretypes:
fe.ngrams(n=5)
if "7gram" in self.featuretypes:
fe.ngrams(n=7)
if "dep" in self.featuretypes:
fe.dependency()
if "srl" in self.featuretypes:
fe.srl()
if "ne" in self.featuretypes:
fe.ne()
if "errorprob" in self.featuretypes:
pass
if "topic" in self.featuretypes:
pass
augf = proc_easyadapt(fe.features, domain=domain)
assert augf and _labelid and v
_flist.append(augf)
_labellist_int.append(_labelid)
_labellist_str.append(v)
except ValueError:
logging.debug(pformat("CaseMaker feature extraction: couldn't find the verb"))
except:
print v
# else:
# _flist.append(self.nullfeature)
# _labellist_int.append(_labelid)
# _labellist_str.append(v)
return _flist, _labellist_str, _labellist_int
def get_features(tags=[], v="", v_idx=None, features=[]):
fe = SentenceFeatures(tags=tags, verb=v, v_idx=v_idx)
if "chunk" in features:
fe.chunk()
if "3gram" in features:
fe.ngrams(n=3)
if "5gram" in features:
fe.ngrams(n=5)
if "7gram" in features:
fe.ngrams(n=7)
if "dependency" in features:
fe.dependency()
if "ne" in features:
fe.ne()
if "srl" in features:
fe.srl()
if "topic" in features:
fe.topic()
if "errorprob" in features:
fe.ep()
# print pformat(fe.features)
return proc_easyadapt(fe.features, domain="tgt")