def print_sparql_queries():
import argparse
parser = argparse.ArgumentParser(description="Dump qa entity pairs.")
parser.add_argument("--config",
default="config.cfg",
help="The configuration file to use.")
parser.add_argument("--output",
help="The file to dump results to.")
args = parser.parse_args()
globals.read_configuration(args.config)
scorer_globals.init()
parameters = translator.TranslatorParameters()
parameters.require_relation_match = False
parameters.restrict_answer_type = False
dataset = "webquestions_test_filter"
sparql_backend = globals.get_sparql_backend(globals.config)
queries = get_evaluated_queries(dataset, True, parameters)
for index, query in enumerate(queries):
print "--------------------------------------------"
print query.utterance
print "\n".join([str((entity.__class__, entity.entity)) for entity in query.eval_candidates[0].query_candidate.query.identified_entities])
for eval_candidate in query.eval_candidates:
query_candidate = eval_candidate.query_candidate
query_candidate.sparql_backend = sparql_backend
notable_types = query_candidate.get_answers_notable_types()
if notable_types:
print notable_types
print query_candidate.graph_as_simple_string().encode("utf-8")
print query_candidate.to_sparql_query().encode("utf-8")
print "\n\n"
def get_number_of_external_entities():
import scorer_globals
globals.read_configuration('config_webentity.cfg')
parser = CoreNLPParser.init_from_config()
entity_linker = WebSearchResultsExtenderEntityLinker.init_from_config()
entity_linker.topn_entities = 100000
scorer_globals.init()
parameters = translator.TranslatorParameters()
parameters.require_relation_match = False
parameters.restrict_answer_type = False
datasets = ["webquestions_split_train", "webquestions_split_dev",]
# datasets = ["webquestions_split_train_externalentities", "webquestions_split_dev_externalentities",]
# datasets = ["webquestions_split_train_externalentities3", "webquestions_split_dev_externalentities3",]
external_entities_count = []
for dataset in datasets:
queries = load_eval_queries(dataset)
for index, query in enumerate(queries):
entities = entity_linker.identify_entities_in_tokens(parser.parse(query.utterance).tokens, text=query.utterance, find_dates=False)
print "-------------------------"
print query.utterance
print "\n".join(map(str, sorted(entities, key=lambda entity: entity.external_entity_count, reverse=True)))
external_entities_count.append(0)
for entity in entities:
if entity.external_entity:
external_entities_count[-1] += 1
if index % 100 == 0:
print >> sys.stderr, "%s queries processed" % index
print "========================================="
print external_entities_count
print sum(external_entities_count)
print len(external_entities_count)
def train_type_model():
globals.read_configuration('config.cfg')
parser = globals.get_parser()
scorer_globals.init()
datasets = ["webquestions_split_train", ]
parameters = translator.TranslatorParameters()
parameters.require_relation_match = False
parameters.restrict_answer_type = False
feature_extractor = FeatureExtractor(False, False, n_gram_types_features=True)
features = []
labels = []
for dataset in datasets:
queries = get_evaluated_queries(dataset, True, parameters)
for index, query in enumerate(queries):
tokens = [token.lemma for token in parser.parse(query.utterance).tokens]
n_grams = get_grams_feats(tokens)
answer_entities = [mid for answer in query.target_result
for mid in KBEntity.get_entityid_by_name(answer, keep_most_triples=True)]
correct_notable_types = set(filter(lambda x: x,
[KBEntity.get_notable_type(entity_mid)
for entity_mid in answer_entities]))
other_notable_types = set()
for candidate in query.eval_candidates:
entities = [mid for entity_name in candidate.prediction
for mid in KBEntity.get_entityid_by_name(entity_name, keep_most_triples=True)]
other_notable_types.update(set([KBEntity.get_notable_type(entity_mid) for entity_mid in entities]))
incorrect_notable_types = other_notable_types.difference(correct_notable_types)
for type in correct_notable_types.union(incorrect_notable_types):
if type in correct_notable_types:
labels.append(1)
else:
labels.append(0)
features.append(feature_extractor.extract_ngram_features(n_grams, [type, ], "type"))
with open("type_model_data.pickle", 'wb') as out:
pickle.dump((features, labels), out)
label_encoder = LabelEncoder()
labels = label_encoder.fit_transform(labels)
vec = DictVectorizer(sparse=True)
X = vec.fit_transform(features)
feature_selector = SelectPercentile(chi2, percentile=5).fit(X, labels)
vec.restrict(feature_selector.get_support())
X = feature_selector.transform(X)
type_scorer = SGDClassifier(loss='log', class_weight='auto',
n_iter=1000,
alpha=1.0,
random_state=999,
verbose=5)
type_scorer.fit(X, labels)
with open("type-model.pickle", 'wb') as out:
pickle.dump((vec, type_scorer), out)
def get_question_terms():
import scorer_globals
globals.read_configuration('config_webentity.cfg')
scorer_globals.init()
datasets = ["webquestionstrain", "webquestionstest",]
question_tokens = set()
for dataset in datasets:
queries = load_eval_queries(dataset)
for index, query in enumerate(queries):
question_tokens.update(token for token in tokenize(query.utterance))
print question_tokens
def main():
import argparse
parser = argparse.ArgumentParser(description="Console based translation.")
parser.add_argument("ranker_name",
default="WQ_Ranker",
help="The ranker to use.")
parser.add_argument("--config",
default="config.cfg",
help="The configuration file to use.")
args = parser.parse_args()
globals.read_configuration(args.config)
scorer_globals.init()
if args.ranker_name not in scorer_globals.scorers_dict:
logger.error("%s is not a valid ranker" % args.ranker_name)
logger.error("Valid rankers are: %s " % (" ".join(scorer_globals.scorers_dict.keys())))
logger.info("Using ranker %s" % args.ranker_name)
ranker = scorer_globals.scorers_dict[args.ranker_name]
translator = SparqlQueryTranslator.init_from_config()
translator.set_scorer(ranker)
while True:
try:
sys.stdout.write("enter question> ")
sys.stdout.flush()
query = sys.stdin.readline().strip()
logger.info("Translating query: %s" % query)
results = translator.translate_and_execute_query(query)
logger.info("Done translating query: %s" % query)
logger.info("#candidates: %s" % len(results))
logger.info("------------------- Candidate features ------------------")
for rank, result in enumerate(results[:10]):
logger.info("RANK " + str(rank))
logger.info(result.query_candidate.relations)
logger.info(result.query_candidate.get_results_text())
if result.query_candidate.features:
logger.info("Features: " + str(result.query_candidate.features))
logger.info("---------------------------------------------------------")
if len(results) > 0:
best_candidate = results[0].query_candidate
sparql_query = best_candidate.to_sparql_query()
result_rows = results[0].query_result_rows
result = []
# Usually we get a name + mid.
for r in result_rows:
if len(r) > 1:
result.append("%s (%s)" % (r[1], r[0]))
else:
result.append("%s" % r[0])
logger.info("SPARQL query: %s" % sparql_query)
logger.info("Result: %s " % " ".join(result))
except Exception as e:
logger.error(e.message)
if __name__ == "__main__":
# print_sparql_queries()
# exit()
import argparse
parser = argparse.ArgumentParser(description="Dump qa entity pairs.")
parser.add_argument("--config",
default="config.cfg",
help="The configuration file to use.")
parser.add_argument("--output",
help="The file to dump results to.")
args = parser.parse_args()
globals.read_configuration(args.config)
scorer_globals.init()
parameters = translator.TranslatorParameters()
parameters.require_relation_match = False
parameters.restrict_answer_type = False
# datasets = ["webquestions_split_train", "webquestions_split_dev",]
# datasets = ["webquestions_split_train_externalentities", "webquestions_split_dev_externalentities",]
# datasets = ["webquestions_split_train_externalentities3", "webquestions_split_dev_externalentities3",]
datasets = ["webquestions_train_externalentities_all", "webquestions_test_externalentities_all", ]
count = 0
correct_relations = set()
positions = []
for dataset in datasets:
queries = get_evaluated_queries(dataset, True, parameters)
def extract_npmi_ngram_type_pairs():
globals.read_configuration('config.cfg')
scorer_globals.init()
datasets = ["webquestions_split_train", ]
parameters = translator.TranslatorParameters()
parameters.require_relation_match = False
parameters.restrict_answer_type = False
n_gram_type_counts = dict()
type_counts = dict()
n_gram_counts = dict()
total = 0
year_pattern = re.compile("[0-9]+")
for dataset in datasets:
queries = get_evaluated_queries(dataset, True, parameters)
for index, query in enumerate(queries):
if query.oracle_position != -1 and query.oracle_position <= len(query.eval_candidates):
correct_candidate = query.eval_candidates[query.oracle_position - 1]
logger.info(query.utterance)
logger.info(correct_candidate.query_candidate)
n_grams = set(get_n_grams_features(correct_candidate.query_candidate))
answer_entities = [mid for answer in query.target_result
if year_pattern.match(answer) is None
for mid in KBEntity.get_entityid_by_name(answer, keep_most_triples=True)]
correct_notable_types = set(filter(lambda x: x,
[KBEntity.get_notable_type(entity_mid)
for entity_mid in answer_entities]))
for notable_type in correct_notable_types:
if notable_type not in type_counts:
type_counts[notable_type] = 0
type_counts[notable_type] += 1
for n_gram in n_grams:
if n_gram not in n_gram_counts:
n_gram_counts[n_gram] = 0
n_gram_counts[n_gram] += 1
for notable_type in correct_notable_types:
pair = (n_gram, notable_type)
if pair not in n_gram_type_counts:
n_gram_type_counts[pair] = 0
n_gram_type_counts[pair] += 1
total += 1
npmi = dict()
from math import log
for n_gram_type_pair, n_gram_type_count in n_gram_type_counts.iteritems():
if n_gram_type_count > 4:
n_gram, type = n_gram_type_pair
npmi[n_gram_type_pair] = (log(n_gram_type_count) - log(n_gram_counts[n_gram]) - log(type_counts[type]) +
log(total)) / (-log(n_gram_type_count) + log(total))
with open("type_model_npmi.pickle", 'wb') as out:
pickle.dump(npmi, out)
import operator
npmi = sorted(npmi.items(), key=operator.itemgetter(1), reverse=True)
print "\n".join(map(str, npmi[:50]))
def main():
import argparse
parser = argparse.ArgumentParser(description='Learn or test a'
' scorer model.')
parser.add_argument('--no-cached',
default=False,
action='store_true',
help='Don\'t use cached data if available.')
parser.add_argument('--config',
default='config.cfg',
help='The configuration file to use.')
subparsers = parser.add_subparsers(help='command help')
train_parser = subparsers.add_parser('train', help='Train a scorer.')
train_parser.add_argument('scorer_name',
help='The scorer to train.')
train_parser.set_defaults(which='train')
test_parser = subparsers.add_parser('test', help='Test a scorer.')
test_parser.add_argument('scorer_name',
help='The scorer to test.')
test_parser.add_argument('test_dataset',
help='The dataset on which to test the scorer.')
test_parser.add_argument('--avg_runs',
type=int,
default=1,
help='Over how many runs to average.')
test_parser.set_defaults(which='test')
traintest_parser = subparsers.add_parser('traintest', help='Train and test a scorer.')
traintest_parser.add_argument('scorer_name',
help='The scorer to test.')
traintest_parser.add_argument('test_dataset',
help='The dataset on which to test the scorer.')
traintest_parser.add_argument('--avg_runs',
type=int,
default=1,
help='Over how many runs to average.')
traintest_parser.set_defaults(which='traintest')
cv_parser = subparsers.add_parser('cv', help='Cross-validate a scorer.')
cv_parser.add_argument('scorer_name',
help='The scorer to test.')
cv_parser.add_argument('dataset',
help='The dataset on which to compute cv scores.')
cv_parser.add_argument('--n_folds',
type=int,
default=6,
help='The number of folds.')
cv_parser.add_argument('--avg_runs',
type=int,
default=1,
help='Over how many runs to average.')
cv_parser.set_defaults(which='cv')
print_parser = subparsers.add_parser('print', help='Print ranked results.')
print_parser.add_argument('scorer_name',
help='The scorer to test.')
print_parser.add_argument('dataset',
help='The dataset on which to compute cv scores.')
print_parser.set_defaults(which='print')
args = parser.parse_args()
# Read global config.
globals.read_configuration(args.config)
scorer_globals.init()
# Fix randomness.
random.seed(999)
use_cache = not args.no_cached
if args.which == 'train':
train(args.scorer_name, use_cache)
elif args.which == 'test':
test(args.scorer_name, args.test_dataset, use_cache,
avg_runs=args.avg_runs)
elif args.which == 'traintest':
train(args.scorer_name, use_cache)
test(args.scorer_name, args.test_dataset, use_cache,
avg_runs=args.avg_runs)
elif args.which == 'cv':
cv(args.scorer_name, args.dataset, use_cache, n_folds=args.n_folds,
avg_runs=args.avg_runs)
elif args.which == 'print':
eval_print(args.scorer_name, args.dataset, use_cache)