def generate_sentence_types(ttk, sen, words):
for fname in os.listdir(ttk):
if not fname.endswith('.lif'):
continue
print("{} ... ".format(os.path.basename(fname)))
if DEBUG:
GOOD.write(">>> %s\n>>> %s\n>>> %s\n\n" %
('-' * 100, fname, '-' * 100))
BAD.write(">>> %s\n>>> %s\n>>> %s\n\n" %
('-' * 100, fname, '-' * 100))
fname_in = os.path.join(ttk, fname)
fname_out = os.path.join(sen, fname)
lif_in = LIF(fname_in)
lif_out = LIF(json_object=lif_in.as_json())
sentences_view = _create_view()
lif_out.views = [sentences_view]
good_sentences = 0
bad_sentences = 0
view = lif_in.get_view('v1')
for anno in view.annotations:
if anno.type.endswith('Sentence'):
sc = SentenceClassifier(lif_in, anno, words)
if sc.is_crap():
if DEBUG:
BAD.write(">>> %f\n%s\n\n" % (sc.ratio, sc.text))
anno.features['type'] = 'crap'
bad_sentences += 1
else:
if DEBUG:
GOOD.write(">>> %f\n%s\n\n" % (sc.ratio, sc.text))
anno.features['type'] = 'normal'
good_sentences += 1
sentences_view.annotations.append(anno)
if DEBUG:
print(" (good={:d} bad={:d})".format(good_sentences,
bad_sentences))
lif_out.write(fname=fname_out, pretty=True)
#break
print
class Document(object):
def __init__(self, fname, data_dir, lif_file, top_file, har_file):
"""Build a single LIF object with all relevant annotations. The annotations
themselves are stored in the Annotations object in self.annotations."""
self.id = fname
self.fname = fname
self.data_dir = data_dir
self.lif = LIF(json_file=lif_file)
self.top = LIF(json_file=top_file)
self.har = LIF(json_file=har_file)
# NOTE: no idea why this was needed
# TODO: there is an error in lif.py in line 80 where the json object is
# handed in as the id
fix_view('doc', self.lif.views[0])
fix_view('top', self.top.views[0])
self.lif.views.append(self.top.views[0])
self.annotations = Annotations(self.id,
fname,
doc=self,
text=self.lif.text.value)
self.annotations.text = self.lif.text.value
self._collect_authors()
self._collect_topics()
self._collect_relations()
def get_view(self, identifier):
return self.lif.get_view(identifier)
def _collect_authors(self):
"""Just get the authors from the metadata and put them in the index."""
def okay(a):
# need to do this because the filter in covid.py is faulty
return len(a) > 3 and not a[0] == ' ' and not a[-1] == ' '
self.annotations.authors = [
a for a in self.lif.metadata['authors'] if okay(a)
]
def _collect_topics(self):
"""Collect the topics and put them on a list in the index."""
view = self.get_view("top")
for annotation in view.annotations:
if annotation.type.endswith('SemanticTag'):
topic_name = annotation.features['topic_name']
self.annotations.topics.append(topic_name)
for topic_element in topic_name.split():
self.annotations.topic_elements.append(topic_element)
self.annotations.topic_elements = sorted(
set(self.annotations.topic_elements))
def _collect_relations(self):
added = False
for relobj, subjs in self.har.metadata['relations'].items():
self.annotations.containers.append(relobj)
self.annotations.proteins.append(relobj.rsplit('-', 2)[0])
for subj in subjs:
self.annotations.proteins.append(subj)
if relobj in self.annotations.relations:
added = True
self.annotations.relations[relobj].append(subj)
# print(self.annotations.proteins)
# print(self.annotations.containers)
# if added:
# print(self.annotations.relations)
def write(self, dirname):
self.annotations.write(os.path.join(dirname, self.fname),
self.lif.metadata["year"])
def pp(self, prefix=''):
views = ["%s:%d" % (view.id, len(view)) for view in self.lif.views]
print("%s<Document id=%s '%s'>" % (prefix, self.id, self.fname))
print(" <Views %s>" % ' '.join(views))
print(" %s\n" % self.annotations)
