def main():
start_time = time.time()
parser = argparse.ArgumentParser(description='')
parser.add_argument("actions",
default="classify",
help="Actions to be performed.")
parser.add_argument(
"--goldstd",
default="",
dest="goldstd",
nargs="+",
help="Gold standard to be used. Will override corpus, annotations",
choices=config.paths.keys())
parser.add_argument("--submodels",
default="",
nargs='+',
help="sub types of classifiers"),
parser.add_argument("--models",
dest="models",
help="model destination path, without extension")
parser.add_argument("--pairtype",
dest="ptype",
help="type of pairs to be considered",
default="all")
parser.add_argument("--doctype",
dest="doctype",
help="type of document to be considered",
default="all")
parser.add_argument(
"-o",
"--output",
"--format",
dest="output",
nargs=2,
help="format path; output formats: xml, html, tsv, text, chemdner.")
parser.add_argument("--log",
action="store",
dest="loglevel",
default="WARNING",
help="Log level")
parser.add_argument("--kernel",
action="store",
dest="kernel",
default="svmtk",
help="Kernel for relation extraction")
options = parser.parse_args()
# set logger
numeric_level = getattr(logging, options.loglevel.upper(), None)
if not isinstance(numeric_level, int):
raise ValueError('Invalid log level: %s' % options.loglevel)
while len(logging.root.handlers) > 0:
logging.root.removeHandler(logging.root.handlers[-1])
logging_format = '%(asctime)s %(levelname)s %(filename)s:%(lineno)s:%(funcName)s %(message)s'
logging.basicConfig(level=numeric_level, format=logging_format)
logging.getLogger().setLevel(numeric_level)
logging.getLogger("requests.packages").setLevel(30)
logging.info("Processing action {0} on {1}".format(options.actions,
options.goldstd))
# set configuration variables based on the goldstd option if the corpus has a gold standard,
# or on corpus and annotation options
# pre-processing options
if options.actions == "load_corpus":
if len(options.goldstd) > 1:
print "load only one corpus each time"
sys.exit()
options.goldstd = options.goldstd[0]
corpus_format = config.paths[options.goldstd]["format"]
corpus_path = config.paths[options.goldstd]["text"]
corpus_ann = config.paths[options.goldstd]["annotations"]
corenlp_client = StanfordCoreNLP('http://localhost:9000')
# corpus = load_corpus(options.goldstd, corpus_path, corpus_format, corenlp_client)
corpus = SeeDevCorpus(corpus_path)
corpus.load_corpus(corenlp_client)
corpus.save(config.paths[options.goldstd]["corpus"])
if corpus_ann: #add annotation if it is not a test set
corpus.load_annotations(corpus_ann, "all")
corpus.save(config.paths[options.goldstd]["corpus"])
elif options.actions == "annotate": # rext-add annotation to corpus
if len(options.goldstd) > 1:
print "load only one corpus each time"
sys.exit()
options.goldstd = options.goldstd[0]
corpus_path = config.paths[options.goldstd]["corpus"]
corpus_ann = config.paths[options.goldstd]["annotations"]
logging.info("loading corpus %s" % corpus_path)
corpus = pickle.load(open(corpus_path, 'rb'))
logging.debug("loading annotations...")
# corpus.clear_annotations("all")
corpus.load_annotations(corpus_ann, "all", options.ptype)
# corpus.get_invalid_sentences()
corpus.save(config.paths[options.goldstd]["corpus"])
else:
#corpus = SeeDevCorpus("corpus/" + "&".join(options.goldstd))
corpus_path = config.paths[options.goldstd[0]]["corpus"]
logging.info("loading corpus %s" % corpus_path)
corpus = pickle.load(open(corpus_path, 'rb'))
if options.actions == "add_sentences":
corpus.add_more_sentences(options.models)
elif options.actions == "train_relations":
if options.ptype == "all":
ptypes = config.pair_types.keys()
# ptypes = config.event_types.keys()
else:
ptypes = [options.ptype]
for p in ptypes:
print p
if options.kernel == "jsre":
model = JSREKernel(corpus, p, train=True)
elif options.kernel == "svmtk":
model = SVMTKernel(corpus, p)
elif options.kernel == "stanfordre":
model = StanfordRE(corpus, p)
elif options.kernel == "multir":
model = MultiR(corpus, p)
elif options.kernel == "scikit":
model = ScikitRE(corpus, p)
elif options.kernel == "crf":
model = CrfSuiteRE(corpus, p)
model.train()
# testing
elif options.actions == "test_relations":
if options.ptype == "all":
ptypes = config.pair_types.keys()
# ptypes = config.event_types.keys()
all_results = ResultsRE(options.output[1])
all_results.corpus = corpus
all_results.path = options.output[1]
else:
ptypes = [options.ptype]
for p in ptypes:
print p
if options.kernel == "jsre":
model = JSREKernel(corpus, p, train=False)
elif options.kernel == "svmtk":
model = SVMTKernel(corpus, p)
elif options.kernel == "rules":
model = RuleClassifier(corpus, p)
elif options.kernel == "stanfordre":
model = StanfordRE(corpus, p)
elif options.kernel == "scikit":
model = ScikitRE(corpus, p)
elif options.kernel == "crf":
model = CrfSuiteRE(corpus, p, test=True)
model.load_classifier()
model.test()
results = model.get_predictions(corpus)
# results.save(options.output[1] + "_" + p.lower() + ".pickle")
# results.load_corpus(options.goldstd[0])
results.path = options.output[1] + "_" + p.lower()
goldset = get_gold_ann_set(
config.paths[options.goldstd[0]]["format"],
config.paths[options.goldstd[0]]["annotations"], "all", p,
config.paths[options.goldstd[0]]["text"])
get_relations_results(results, options.models, goldset[1], [],
[])
if options.ptype == "all":
for did in results.document_pairs:
if did not in all_results.document_pairs:
all_results.document_pairs[did] = Pairs(did=did)
all_results.document_pairs[
did].pairs += results.document_pairs[did].pairs
if options.ptype == "all":
goldset = get_gold_ann_set(
config.paths[options.goldstd[0]]["format"],
config.paths[options.goldstd[0]]["annotations"], "all",
"all", config.paths[options.goldstd[0]]["text"])
get_relations_results(all_results, options.models, goldset[1],
[], [])
write_seedev_results(all_results, options.output[1])
elif options.actions == "train_sentences": #and evaluate
if options.ptype == "all":
avg = [0, 0, 0]
for p in config.pair_types:
print p
tps, fps, fns = corpus.train_sentence_classifier(p)
if tps == 0 and fns == 0:
precision, recall, fscore = 0, 1, 1
else:
precision = 1.0 * tps / (fps + tps)
recall = 1.0 * fns / (fns + tps)
fscore = 2.0 * precision * recall / (recall +
precision)
print precision, recall, fscore
avg[0] += tps
avg[1] += fps
avg[2] += fns
#print [a/len(config.pair_types) for a in avg]
precision = 1.0 * avg[1] / (avg[0] + avg[1])
recall = 1.0 * avg[2] / (avg[0] + avg[2])
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
else:
res = corpus.train_sentence_classifier(options.ptype)
print res
corpus.save(config.paths[options.goldstd[0]]["corpus"])
elif options.actions == "test_sentences": #and evaluate
if options.ptype == "all":
avg = [0, 0, 0]
for p in config.pair_types:
print p
tps, fps, fns = corpus.test_sentence_classifier(p)
if tps == 0 and fns == 0:
precision, recall, fscore = 0, 1, 1
else:
precision = 1.0 * tps / (fps + tps)
recall = 1.0 * fns / (fns + tps)
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
avg[0] += tps
avg[1] += fps
avg[2] += fns
#print [a/len(config.pair_types) for a in avg]
precision = 1.0 * avg[1] / (avg[0] + avg[1])
recall = 1.0 * avg[2] / (avg[0] + avg[2])
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
else:
res = corpus.test_sentence_classifier(options.ptype)
print res
corpus.save(config.paths[options.goldstd[0]]["corpus"])
total_time = time.time() - start_time
logging.info("Total time: %ss" % total_time)
def main():
start_time = time.time()
parser = argparse.ArgumentParser(description='')
parser.add_argument("--train",
nargs="+",
help="Gold standards to be used for training")
parser.add_argument("--test",
nargs="+",
help="Gold standards to be used for testing")
parser.add_argument("--tag",
dest="tag",
default="0",
help="Tag to identify the experiment")
parser.add_argument("--emodels",
dest="emodels",
help="model destination path, without extension",
nargs="+")
parser.add_argument("--rmodels",
dest="rmodels",
help="model destination path, without extension")
parser.add_argument("--entitytype",
dest="etype",
help="type of entities to be considered",
default="all")
parser.add_argument("--pairtype",
dest="ptype",
help="type of pairs to be considered",
default="all")
parser.add_argument("--crf",
dest="crf",
help="CRF implementation",
default="stanford",
choices=["stanford", "crfsuite", "banner"])
parser.add_argument("--log",
action="store",
dest="loglevel",
default="WARNING",
help="Log level")
parser.add_argument("--rules",
default=[],
nargs='+',
help="aditional post processing rules")
parser.add_argument("--kernel",
action="store",
dest="kernel",
default="svmtk",
help="Kernel for relation extraction")
parser.add_argument("--results",
dest="results",
help="Results object pickle.")
options = parser.parse_args()
# set logger
numeric_level = getattr(logging, options.loglevel.upper(), None)
if not isinstance(numeric_level, int):
raise ValueError('Invalid log level: %s' % options.loglevel)
while len(logging.root.handlers) > 0:
logging.root.removeHandler(logging.root.handlers[-1])
logging_format = '%(asctime)s %(levelname)s %(filename)s:%(lineno)s:%(funcName)s %(message)s'
logging.basicConfig(level=numeric_level, format=logging_format)
logging.getLogger().setLevel(numeric_level)
logging.getLogger("requests.packages").setLevel(30)
# logging.info("Processing action {0} on {1}".format(options.actions, options.goldstd))
relations = set()
with open("corpora/transmir/transmir_relations.txt") as rfile:
for l in rfile:
relations.add(tuple(l.strip().split('\t')))
with open("temp/mil.train", 'w') as f:
f.write("")
# train_corpus = Corpus("corpus/" + "&".join(options.goldstd[0]))
total_entities = 0
for goldstd in options.train:
corpus_path = paths[goldstd]["corpus"]
logging.info("loading corpus %s" % corpus_path)
train_corpus = pickle.load(open(corpus_path, 'rb'))
for sentence in train_corpus.get_sentences(options.emodels[0]):
for e in sentence.entities.elist[options.emodels[0]]:
if e.normalized_score > 0:
total_entities += 1
# with open("mirna_ds-pmids.txt", 'w') as pmidfile:
# for did in train_corpus.documents:
# pmidfile.write(did + "\n")
train_model = MILClassifier(train_corpus,
options.ptype,
relations,
ner=options.emodels[0])
train_model.load_kb("corpora/transmir/transmir_relations.txt")
train_model.generateMILdata(test=False)
train_model.write_to_file("temp/mil.train")
train_model = None
train_corpus = None
print "total entities:", total_entities
train_model = MILClassifier(None,
options.ptype,
relations,
ner=options.emodels[0],
generate=False,
modelname=options.rmodels)
train_model.load_kb("corpora/transmir/transmir_relations.txt")
train_model.load_from_file("temp/mil.train")
#train_model.generateMILdata(test=False)
train_model.train()
# test_corpus = Corpus("corpus/" + "&".join(options.goldstd[1]))
test_sets = []
for g in options.test:
corpus_path = paths[g]["corpus"]
logging.info("loading corpus %s" % corpus_path)
test_corpus = pickle.load(open(corpus_path, 'rb'))
test_sets.append(test_corpus)
for i, test_corpus in enumerate(test_sets):
logging.info("evaluation {}".format(options.test[i]))
test_model = MILClassifier(test_corpus,
options.ptype,
relations,
test=True,
ner=options.emodels[i + 1],
modelname=options.rmodels)
test_model.load_kb("corpora/transmir/transmir_relations.txt")
test_model.generateMILdata(test=False)
test_model.load_classifier()
#test_model.vectorizer = train_model.vectorizer
#test_model.classifier = train_model.classifier
test_model.test()
results = test_model.get_predictions(test_corpus)
results.path = options.results + "-" + options.test[i]
results.save(options.results + "-" + options.test[i] + ".pickle")
results.load_corpus(options.test[i])
if options.test[i] != "mirna_cf_annotated":
logging.info("loading gold standard %s" %
paths[options.test[i]]["annotations"])
goldset = get_gold_ann_set(paths[options.test[i]]["format"],
paths[options.test[i]]["annotations"],
options.etype, options.ptype,
paths[options.test[i]]["text"])
if options.test[i] in ("transmir_annotated", "miRTex_test"):
get_list_results(results,
options.kernel,
goldset[1], {},
options.rules,
mode="re")
else:
get_relations_results(results, options.kernel, goldset[1], {},
options.rules)
else:
total_entities = 0
for sentence in test_corpus.get_sentences(options.emodels[1]):
for e in sentence.entities.elist[options.emodels[1]]:
if e.normalized_score > 0:
total_entities += 1
print "total entities:", total_entities
#sysresults = results.corpus.get_unique_results(options.kernel, {}, options.rules, mode="re")
sysresults = []
for did in results.corpus.documents:
for p in results.corpus.documents[did].pairs.pairs:
if options.kernel in p.recognized_by:
sentence = results.corpus.documents[did].get_sentence(
p.entities[0].sid)
# print p.entities[0].text, p.entities[1].text
sysresults.append(
(p.entities[0].sid, p.entities[0].normalized,
p.entities[1].normalized,
"{}=>{}".format(p.entities[0].normalized,
p.entities[1].normalized),
p.recognized_by[options.kernel], sentence.text))
rels = {}
for x in sysresults:
pair = (x[1], x[2])
if pair not in rels:
rels[pair] = []
#print x[0], x[-1], x[3]
stext = x[0] + ": " + x[-1] + " (" + x[3] + ")"
# stext = ""
add = True
for i in rels[pair]:
if i[0].startswith(x[0]):
add = False
break
if add:
rels[pair].append((stext, x[4]))
# for t in rels.items():
# print t
o = OrderedDict(
sorted(rels.items(), key=lambda t: t[1][0][1], reverse=True))
for x in o:
print x, len(o[x])
for s in o[x]:
print "\t", s[0].encode("utf-8")
print
for x in o:
conf = round(o[x][0][1], 3)
unique_dids = set([sid[0].split(".")[0] for sid in o[x]])
# print unique_dids
print x[0] + "\t" + x[1] + "\t" + str(len(o[x])) + "\t" + str(
len(unique_dids)) + "\t" + str(conf)
#max_width = table_instance.column_max_width(2)
#for i, line in enumerate(table_instance.table_data):
# wrapped_string = '\n'.join(wrap(line[2], max_width))
# table_instance.table_data[i][2] = wrapped_string
total_time = time.time() - start_time
print "Total time: %ss" % total_time
def main():
start_time = time.time()
parser = argparse.ArgumentParser(description='')
parser.add_argument("actions", default="classify", help="Actions to be performed.")
parser.add_argument("--goldstd", default="", dest="goldstd", nargs="+",
help="Gold standard to be used. Will override corpus, annotations",
choices=paths.keys())
parser.add_argument("--submodels", default="", nargs='+', help="sub types of classifiers"),
parser.add_argument("--models", dest="models", help="model destination path, without extension")
parser.add_argument("--pairtype", dest="ptype", help="type of pairs to be considered", default="all")
parser.add_argument("--doctype", dest="doctype", help="type of document to be considered", default="all")
parser.add_argument("-o", "--output", "--format", dest="output",
nargs=2, help="format path; output formats: xml, html, tsv, text, chemdner.")
parser.add_argument("--log", action="store", dest="loglevel", default="WARNING", help="Log level")
parser.add_argument("--kernel", action="store", dest="kernel", default="svmtk", help="Kernel for relation extraction")
options = parser.parse_args()
# set logger
numeric_level = getattr(logging, options.loglevel.upper(), None)
if not isinstance(numeric_level, int):
raise ValueError('Invalid log level: %s' % options.loglevel)
while len(logging.root.handlers) > 0:
logging.root.removeHandler(logging.root.handlers[-1])
logging_format = '%(asctime)s %(levelname)s %(filename)s:%(lineno)s:%(funcName)s %(message)s'
logging.basicConfig(level=numeric_level, format=logging_format)
logging.getLogger().setLevel(numeric_level)
logging.getLogger("requests.packages").setLevel(30)
logging.info("Processing action {0} on {1}".format(options.actions, options.goldstd))
# set configuration variables based on the goldstd option if the corpus has a gold standard,
# or on corpus and annotation options
# pre-processing options
if options.actions == "load_corpus":
if len(options.goldstd) > 1:
print "load only one corpus each time"
sys.exit()
options.goldstd = options.goldstd[0]
corpus_format = paths[options.goldstd]["format"]
corpus_path = paths[options.goldstd]["text"]
corpus_ann = paths[options.goldstd]["annotations"]
corenlp_client = StanfordCoreNLP('http://localhost:9000')
# corpus = load_corpus(options.goldstd, corpus_path, corpus_format, corenlp_client)
corpus = SeeDevCorpus(corpus_path)
corpus.load_corpus(corenlp_client)
corpus.save(paths[options.goldstd]["corpus"])
if corpus_ann: #add annotation if it is not a test set
corpus.load_annotations(corpus_ann, "all")
corpus.save(paths[options.goldstd]["corpus"])
elif options.actions == "annotate": # rext-add annotation to corpus
if len(options.goldstd) > 1:
print "load only one corpus each time"
sys.exit()
options.goldstd = options.goldstd[0]
corpus_path = paths[options.goldstd]["corpus"]
corpus_ann = paths[options.goldstd]["annotations"]
logging.info("loading corpus %s" % corpus_path)
corpus = pickle.load(open(corpus_path, 'rb'))
logging.debug("loading annotations...")
# corpus.clear_annotations("all")
corpus.load_annotations(corpus_ann, "all", options.ptype)
# corpus.get_invalid_sentences()
corpus.save(paths[options.goldstd]["corpus"])
else:
#corpus = SeeDevCorpus("corpus/" + "&".join(options.goldstd))
corpus_path = paths[options.goldstd[0]]["corpus"]
logging.info("loading corpus %s" % corpus_path)
basecorpus = pickle.load(open(corpus_path, 'rb'))
corpus = SeeDevCorpus(corpus_path)
corpus.documents = basecorpus.documents
if options.actions == "add_sentences":
corpus.add_more_sentences(options.models)
elif options.actions == "add_goldstandard":
corpus.convert_entities_to_goldstandard()
corpus.find_ds_relations()
#corpus.save(config.paths[options.goldstd[0]]["corpus"])
elif options.actions == "train_multiple": # Train one classifier for each type of entity in this corpus
# logging.info(corpus.subtypes)
models = TaggerCollection(basepath=options.models, corpus=corpus, subtypes=all_entity_types)
models.train_types()
elif options.actions == "train_relations":
if options.ptype == "all":
ptypes = pair_types.keys()
# ptypes = config.event_types.keys()
else:
ptypes = [options.ptype]
for p in ptypes:
print p
if options.kernel == "jsre":
model = JSREKernel(corpus, p, train=True)
elif options.kernel == "svmtk":
model = SVMTKernel(corpus, p)
elif options.kernel == "stanfordre":
model = StanfordRE(corpus, p)
elif options.kernel == "multir":
model = MultiR(corpus, p)
elif options.kernel == "scikit":
model = ScikitRE(corpus, p)
elif options.kernel == "crf":
model = CrfSuiteRE(corpus, p)
# model.train()
# testing
elif options.actions == "test_multiple":
logging.info("testing with multiple classifiers... {}".format(' '.join(options.submodels)))
models = TaggerCollection(basepath=options.models, subtypes=all_entity_types)
models.load_models()
results = models.test_types(corpus)
final_results = results.combine_results()
logging.info("saving results...")
final_results.save(options.output[1] + ".pickle")
elif options.actions == "test_relations":
if options.ptype == "all":
ptypes = pair_types.keys()
# ptypes = config.event_types.keys()
all_results = ResultsRE(options.output[1])
all_results.corpus = corpus
all_results.path = options.output[1]
else:
ptypes = [options.ptype]
for p in ptypes:
print p
if options.kernel == "jsre":
model = JSREKernel(corpus, p, train=False)
elif options.kernel == "svmtk":
model = SVMTKernel(corpus, p)
elif options.kernel == "rules":
model = RuleClassifier(corpus, p)
elif options.kernel == "stanfordre":
model = StanfordRE(corpus, p)
elif options.kernel == "scikit":
model = ScikitRE(corpus, p)
elif options.kernel == "crf":
model = CrfSuiteRE(corpus, p, test=True)
model.load_classifier()
model.test()
results = model.get_predictions(corpus)
# results.save(options.output[1] + "_" + p.lower() + ".pickle")
# results.load_corpus(options.goldstd[0])
results.path = options.output[1] + "_" + p.lower()
goldset = get_gold_ann_set(paths[options.goldstd[0]]["format"], paths[options.goldstd[0]]["annotations"],
"all", p, paths[options.goldstd[0]]["text"])
get_relations_results(results, options.models, goldset[1],[], [])
if options.ptype == "all":
for did in results.document_pairs:
if did not in all_results.document_pairs:
all_results.document_pairs[did] = Pairs(did=did)
all_results.document_pairs[did].pairs += results.document_pairs[did].pairs
if options.ptype == "all":
goldset = get_gold_ann_set(paths[options.goldstd[0]]["format"], paths[options.goldstd[0]]["annotations"],
"all", "all", paths[options.goldstd[0]]["text"])
get_relations_results(all_results, options.models, goldset[1],[], [])
write_seedev_results(all_results, options.output[1])
elif options.actions == "train_sentences": #and evaluate
if options.ptype == "all":
avg = [0,0,0]
for p in pair_types:
print p
tps, fps, fns = corpus.train_sentence_classifier(p)
if tps == 0 and fns == 0:
precision, recall, fscore = 0, 1, 1
else:
precision = 1.0 * tps / (fps + tps)
recall = 1.0 * fns / (fns + tps)
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
avg[0] += tps
avg[1] += fps
avg[2] += fns
#print [a/len(config.pair_types) for a in avg]
precision = 1.0 * avg[1] / (avg[0] + avg[1])
recall = 1.0 * avg[2] / (avg[0] + avg[2])
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
else:
res = corpus.train_sentence_classifier(options.ptype)
print res
corpus.save(paths[options.goldstd[0]]["corpus"])
elif options.actions == "test_sentences": #and evaluate
if options.ptype == "all":
avg = [0,0,0]
for p in pair_types:
print p
tps, fps, fns = corpus.test_sentence_classifier(p)
if tps == 0 and fns == 0:
precision, recall, fscore = 0, 1, 1
else:
precision = 1.0 * tps / (fps + tps)
recall = 1.0 * fns / (fns + tps)
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
avg[0] += tps
avg[1] += fps
avg[2] += fns
#print [a/len(config.pair_types) for a in avg]
precision = 1.0 * avg[1] / (avg[0] + avg[1])
recall = 1.0 * avg[2] / (avg[0] + avg[2])
fscore = 2.0 * precision * recall / (recall + precision)
print precision, recall, fscore
#else:
# res = corpus.test_sentence_classifier(options.ptype)
# print res
elif options.actions == "evaluate_ner":
if os.path.exists(options.output[1] + ".pickle"):
results = pickle.load(open(options.output[1] + ".pickle", 'rb'))
results.load_corpus(options.goldstd[0])
results.path = options.output[1]
logging.info("loading gold standard %s" % paths[options.goldstd[0]]["annotations"])
for t in all_entity_types:
print t
results.path = options.output[1] + "_" + t
goldset = get_gold_ann_set(paths[options.goldstd[0]]["format"],
paths[options.goldstd[0]]["annotations"],
t, options.ptype, paths[options.goldstd[0]]["text"])
get_results(results, options.models + "_" + t, goldset[0], {}, {})
corpus.save(paths[options.goldstd[0]]["corpus"])
total_time = time.time() - start_time
logging.info("Total time: %ss" % total_time)
def run_crossvalidation(goldstd_list,
corpus,
model,
cv,
crf="stanford",
entity_type="all",
cvlog="cv.log"):
logfile = open(cvlog, 'w')
doclist = corpus.documents.keys()
random.shuffle(doclist)
size = int(len(doclist) / cv)
sublists = chunks(doclist, size)
logging.debug("Chunks:")
logging.debug(sublists)
p, r = [], []
all_results = ResultsNER(model)
all_results.path = model + "_results"
for nlist in range(cv):
testids, trainids = None, None
testids = sublists[nlist]
trainids = list(itertools.chain.from_iterable(sublists[:nlist]))
trainids += list(itertools.chain.from_iterable(sublists[nlist + 1:]))
train_corpus, test_corpus = None, None
print 'CV{} - test set: {}; train set: {}'.format(
nlist, len(testids), len(trainids))
train_corpus = Corpus(
corpus.path + "_train",
documents={did: corpus.documents[did]
for did in trainids})
test_corpus = Corpus(
corpus.path + "_test",
documents={did: corpus.documents[did]
for did in testids})
# logging.debug("train corpus docs: {}".format("\n".join(train_corpus.documents.keys())))
#test_entities = len(test_corpus.get_all_entities("goldstandard"))
#train_entities = len(train_corpus.get_all_entities("goldstandard"))
#logging.info("test set entities: {}; train set entities: {}".format(test_entities, train_entities))
basemodel = model + "_cv{}".format(nlist)
logging.debug('CV{} - test set: {}; train set: {}'.format(
nlist, len(test_corpus.documents), len(train_corpus.documents)))
'''for d in train_corpus.documents:
for s in train_corpus.documents[d].sentences:
print len([t.tags.get("goldstandard") for t in s.tokens if t.tags.get("goldstandard") != "other"])
sys.exit()'''
# train
logging.info('CV{} - TRAIN'.format(nlist))
# train_model = StanfordNERModel(basemodel)
train_model = None
if crf == "stanford":
train_model = StanfordNERModel(basemodel, entity_type)
elif crf == "crfsuite":
train_model = CrfSuiteModel(basemodel, entity_type)
train_model.load_data(train_corpus, feature_extractors.keys())
train_model.train()
# test
logging.info('CV{} - TEST'.format(nlist))
test_model = None
if crf == "stanford":
test_model = StanfordNERModel(basemodel, entity_type)
elif crf == "crfsuite":
test_model = CrfSuiteModel(basemodel, entity_type)
test_model.load_tagger(port=9191 + nlist)
test_model.load_data(test_corpus,
feature_extractors.keys(),
mode="test")
final_results = None
final_results = test_model.test(test_corpus, port=9191 + nlist)
if crf == "stanford":
test_model.kill_process()
final_results.basepath = basemodel + "_results"
final_results.path = basemodel
all_results.entities.update(final_results.entities)
all_results.corpus.documents.update(final_results.corpus.documents)
# validate
"""if config.use_chebi:
logging.info('CV{} - VALIDATE'.format(nlist))
final_results = add_chebi_mappings(final_results, basemodel)
final_results = add_ssm_score(final_results, basemodel)
final_results.combine_results(basemodel, basemodel)"""
# evaluate
logging.info('CV{} - EVALUATE'.format(nlist))
test_goldset = set()
for gs in goldstd_list:
goldset = get_gold_ann_set(config.paths[gs]["format"],
config.paths[gs]["annotations"],
entity_type, "pairtype",
config.paths[gs]["text"])
for g in goldset[0]:
if g[0] in testids:
test_goldset.add(g)
precision, recall = get_results(final_results, basemodel, test_goldset,
{}, [])
# evaluation = run_chemdner_evaluation(config.paths[goldstd]["cem"], basemodel + "_results.txt", "-t")
# values = evaluation.split("\n")[1].split('\t')
p.append(precision)
r.append(recall)
# logging.info("precision: {} recall:{}".format(str(values[13]), str(values[14])))
pavg = sum(p) / cv
ravg = sum(r) / cv
print "precision: average={} all={}".format(
str(pavg), '|'.join([str(pp) for pp in p]))
print "recall: average={} all={}".format(str(ravg),
'|'.join([str(rr) for rr in r]))
all_goldset = set()
for gs in goldstd_list:
goldset = get_gold_ann_set(config.paths[gs]["format"],
config.paths[gs]["annotations"],
entity_type, config.paths[gs]["text"])
for g in goldset:
all_goldset.add(g)
get_results(all_results, model, all_goldset, {}, [])
def run_crossvalidation(goldstd_list, corpus, model, cv, crf="stanford", entity_type="all", cvlog="cv.log"):
logfile = open(cvlog, 'w')
doclist = corpus.documents.keys()
random.shuffle(doclist)
size = int(len(doclist)/cv)
sublists = chunks(doclist, size)
logging.debug("Chunks:")
logging.debug(sublists)
p, r = [], []
all_results = ResultsNER(model)
all_results.path = model + "_results"
for nlist in range(cv):
testids, trainids = None, None
testids = sublists[nlist]
trainids = list(itertools.chain.from_iterable(sublists[:nlist]))
trainids += list(itertools.chain.from_iterable(sublists[nlist+1:]))
train_corpus, test_corpus = None, None
print 'CV{} - test set: {}; train set: {}'.format(nlist, len(testids), len(trainids))
train_corpus = Corpus(corpus.path + "_train", documents={did: corpus.documents[did] for did in trainids})
test_corpus = Corpus(corpus.path + "_test", documents={did: corpus.documents[did] for did in testids})
# logging.debug("train corpus docs: {}".format("\n".join(train_corpus.documents.keys())))
#test_entities = len(test_corpus.get_all_entities("goldstandard"))
#train_entities = len(train_corpus.get_all_entities("goldstandard"))
#logging.info("test set entities: {}; train set entities: {}".format(test_entities, train_entities))
basemodel = model + "_cv{}".format(nlist)
logging.debug('CV{} - test set: {}; train set: {}'.format(nlist, len(test_corpus.documents), len(train_corpus.documents)))
'''for d in train_corpus.documents:
for s in train_corpus.documents[d].sentences:
print len([t.tags.get("goldstandard") for t in s.tokens if t.tags.get("goldstandard") != "other"])
sys.exit()'''
# train
logging.info('CV{} - TRAIN'.format(nlist))
# train_model = StanfordNERModel(basemodel)
train_model = None
if crf == "stanford":
train_model = StanfordNERModel(basemodel, entity_type)
elif crf == "crfsuite":
train_model = CrfSuiteModel(basemodel, entity_type)
train_model.load_data(train_corpus, feature_extractors.keys())
train_model.train()
# test
logging.info('CV{} - TEST'.format(nlist))
test_model = None
if crf == "stanford":
test_model = StanfordNERModel(basemodel, entity_type)
elif crf == "crfsuite":
test_model = CrfSuiteModel(basemodel, entity_type)
test_model.load_tagger(port=9191+nlist)
test_model.load_data(test_corpus, feature_extractors.keys(), mode="test")
final_results = None
final_results = test_model.test(test_corpus, port=9191+nlist)
if crf == "stanford":
test_model.kill_process()
final_results.basepath = basemodel + "_results"
final_results.path = basemodel
all_results.entities.update(final_results.entities)
all_results.corpus.documents.update(final_results.corpus.documents)
# validate
"""if config.use_chebi:
logging.info('CV{} - VALIDATE'.format(nlist))
final_results = add_chebi_mappings(final_results, basemodel)
final_results = add_ssm_score(final_results, basemodel)
final_results.combine_results(basemodel, basemodel)"""
# evaluate
logging.info('CV{} - EVALUATE'.format(nlist))
test_goldset = set()
for gs in goldstd_list:
goldset = get_gold_ann_set(config.corpus_paths.paths[gs]["format"], config.corpus_paths.paths[gs]["annotations"], entity_type,
"pairtype", config.corpus_paths.paths[gs]["text"])
for g in goldset[0]:
if g[0] in testids:
test_goldset.add(g)
precision, recall = get_results(final_results, basemodel, test_goldset, {}, [])
# evaluation = run_chemdner_evaluation(config.paths[goldstd]["cem"], basemodel + "_results.txt", "-t")
# values = evaluation.split("\n")[1].split('\t')
p.append(precision)
r.append(recall)
# logging.info("precision: {} recall:{}".format(str(values[13]), str(values[14])))
pavg = sum(p)/cv
ravg = sum(r)/cv
print "precision: average={} all={}".format(str(pavg), '|'.join([str(pp) for pp in p]))
print "recall: average={} all={}".format(str(ravg), '|'.join([str(rr) for rr in r]))
all_goldset = set()
for gs in goldstd_list:
goldset = get_gold_ann_set(config.corpus_paths.paths[gs]["format"], config.corpus_paths.paths[gs]["annotations"], entity_type, "",
config.corpus_paths.paths[gs]["text"])
for g in goldset[0]:
all_goldset.add(g)
get_results(all_results, model, all_goldset, {}, [])
