def call_cort(text_blob):
mention_features = [
features.fine_type, features.gender, features.number,
features.sem_class, features.deprel, features.head_ner,
features.length, features.head, features.first, features.last,
features.preceding_token, features.next_token, features.governor,
features.ancestry
]
pairwise_features = [
features.exact_match, features.head_match, features.same_speaker,
features.alias, features.sentence_distance, features.embedding,
features.modifier, features.tokens_contained, features.head_contained,
features.token_distance
]
# todo make sure these are exact!
model_abs = '/Users/ryanpanos/Documents/code/cort_experiments/models/model-pair-train+dev.obj' #OMG evil!
# perceptron_path = 'cort.coreference.approaches.mention_ranking.RankingPerceptron'
# extractor_path = ' cort.coreference.approaches.mention_ranking.extract_substructures'
perceptron_path = 'cort.coreference.approaches.mention_ranking.RankingPerceptron'
extractor_path = ' coreference.approaches.mention_ranking.extract_substructures'
corenlp_path = '/Users/ryanpanos/Documents/code/StanfordNLP/stanford-corenlp-full-2016-10-31/' #OMG evil!
clusterer_path = 'cort.coreference.clusterer.all_ante'
# logging.info("Loading model.")
print("Loading model . ... (this takes a while) ")
priors, weights = pickle.load(open(model_abs, "rb"))
print("Model loaded.")
# perceptron = import_helper.import_from_path(perceptron_path)(
# priors=priors,
# weights=weights,
# cost_scaling=0
# )
perceptron = RankingPerceptron(priors=priors,
weights=weights,
cost_scaling=0)
extractor = instance_extractors.InstanceExtractor(
# import_helper.import_from_path(extractor_path),
extract_substructures,
mention_features,
pairwise_features,
cost_functions.null_cost,
perceptron.get_labels())
logging.info("Reading in and preprocessing data.")
p = pipeline.Pipeline(corenlp_path)
testing_corpus = p.run_on_blob("corpus", text_blob)
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,
extractor,
perceptron,
# import_helper.import_from_path(clusterer_path)
all_ante)
testing_corpus.read_coref_decisions(mention_entity_mapping,
antecedent_mapping)
logging.info("Write output to file.")
output_ls = []
for doc in testing_corpus:
output = doc.to_simple_output()
# my_file = codecs.open(doc.identifier + "." + args.suffix, "w", "utf-8")
# my_file.write(output)
print " output: \n" + output
# my_file.close()
output_ls.append(output)
logging.info("Done.")
return
pairwise_features = [
features.exact_match, features.head_match, features.same_speaker,
features.alias, features.sentence_distance, features.embedding,
features.modifier, features.tokens_contained, features.head_contained,
features.token_distance
]
logging.info("Loading model.")
priors, weights = pickle.load(open(args.model, "rb"))
perceptron = import_helper.import_from_path(args.perceptron)(priors=priors,
weights=weights,
cost_scaling=0)
extractor = instance_extractors.InstanceExtractor(
import_helper.import_from_path(args.extractor), mention_features,
pairwise_features, cost_functions.null_cost, perceptron.get_labels())
logging.info("Reading in data.")
testing_corpus = corpora.Corpus.from_file(
"testing", codecs.open(args.input_filename, "r", "utf-8"))
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, extractor, perceptron,
import_helper.import_from_path(args.clusterer))
testing_corpus.read_coref_decisions(mention_entity_mapping, antecedent_mapping)
def __init__(self):
mention_features = [
features.fine_type, features.gender, features.number,
features.sem_class, features.deprel, features.head_ner,
features.length, features.head, features.first, features.last,
features.preceding_token, features.next_token, features.governor,
features.ancestry
]
pairwise_features = [
features.exact_match, features.head_match, features.same_speaker,
features.alias, features.sentence_distance, features.embedding,
features.modifier, features.tokens_contained,
features.head_contained, features.token_distance
]
self.extractor = instance_extractors.InstanceExtractor(
mention_ranking.extract_substructures, mention_features,
pairwise_features, cost_functions.null_cost)
logging.info("Loading model.")
priors, weights = pickle.load(open("latent-model-train.obj", "rb"))
self.perceptron = mention_ranking.RankingPerceptron(priors=priors,
weights=weights,
cost_scaling=0)
logging.info("Loading CoreNLP models.")
self.p = pipeline.Pipeline(
"/home/sebastian/Downloads/stanford-corenlp-full-2015-04-20")
self.root = tki.Tk()
self.root.title("cort Demo")
# create a Frame for the Text and Scrollbar
self.txt_frm = tki.Frame(self.root, width=400, height=200)
self.txt_frm.pack(fill="both", expand=True)
# ensure a consistent GUI size
self.txt_frm.grid_propagate(False)
# implement stretchability
self.txt_frm.grid_rowconfigure(0, weight=1)
self.txt_frm.grid_columnconfigure(0, weight=1)
# create a Text widget
self.txt = tki.Text(self.txt_frm, borderwidth=3, relief="sunken")
self.txt.config(font=("consolas", 12), undo=True, wrap='word')
self.txt.grid(row=0, column=0, sticky="nsew", padx=2, pady=2)
# create a Scrollbar and associate it with txt
scrollb = tki.Scrollbar(self.txt_frm, command=self.txt.yview)
scrollb.grid(row=0, column=1, sticky='nsew')
self.txt['yscrollcommand'] = scrollb.set
self.button = tki.Button(self.root,
text='Resolve Coreference',
command=self.do_coreference)
self.button.pack()