def test_compute_errors(self):
# fake document using a named tuple
document = namedtuple("Document", "annotated_mentions")
doc_gold = document(self.first_cluster + self.second_cluster)
doc_system = document(self.system_cluster)
corpus_gold = corpora.Corpus("fake gold", [doc_gold])
corpus_system = corpora.Corpus("fake system", [doc_system])
ex = error_extractors.ErrorExtractor(
corpus_gold,
spanning_tree_algorithms.recall_closest,
spanning_tree_algorithms.precision_system_output
)
ex.add_system(corpus_system)
self.assertEqual(
data_structures.EnhancedSet([
(self.first_cluster[1], self.first_cluster[0]),
(self.first_cluster[3], self.first_cluster[2]),
(self.first_cluster[5], self.first_cluster[4]),
(self.second_cluster[1], self.second_cluster[0]),
(self.second_cluster[2], self.second_cluster[1]),
]),
ex.get_errors()["fake system"]["recall_errors"]["all"]
)
def run_on_docs(self, identifier, docs):
processed_documents = []
for doc in docs:
processed_documents.append(self.run_on_doc(
codecs.open(doc, "r", "utf-8")
))
return corpora.Corpus(identifier, processed_documents)
def run_on_blob(self, identifier, text_blob):
processed_documents = []
processed_documents.append(self.run_on_doc( # codecs.open(doc, "r", "utf-8")
# identifier,
text_blob,
use_as_blob=True
))
return corpora.Corpus(identifier, processed_documents)
def do_coreference(self):
testing_corpus = corpora.Corpus("input", [
self.p.run_on_doc(io.StringIO(self.txt.get("0.0", tki.END)),
"input")
])
logging.info("Extracting system mentions.")
for doc in testing_corpus:
doc.system_mentions = mention_extractor.extract_system_mentions(
doc)
mention_entity_mapping, antecedent_mapping = experiments.predict(
testing_corpus, self.extractor, self.perceptron,
clusterer.all_ante)
testing_corpus.read_coref_decisions(mention_entity_mapping,
antecedent_mapping)
logging.info("Visualize")
for doc in testing_corpus:
max_id = 0
for mention in doc.system_mentions[1:]:
set_id = mention.attributes["set_id"]
if set_id:
max_id = max(set_id, max_id)
max_id += 1
doc.annotated_mentions = []
for i, mention in enumerate(doc.system_mentions[1:]):
if mention.attributes["set_id"]:
mention.attributes[
"annotated_set_id"] = mention.attributes["set_id"]
else:
mention.attributes["annotated_set_id"] = max_id + i
doc.annotated_mentions.append(mention)
ex = error_extractors.ErrorExtractor(
testing_corpus, spanning_tree_algorithms.recall_accessibility,
spanning_tree_algorithms.precision_system_output)
ex.add_system(testing_corpus)
decisions = ex.get_errors()
visualizer = visualization.Visualizer(decisions,
"input",
for_raw_input=True)
visualizer.run()