def run(EvaluatorClass,
inputCorpusFile,
goldCorpusFile,
parse,
tokenization=None,
target="both",
entityMatchFunction=compareEntitiesSimple,
removeIntersentenceInteractions=False):
print >> sys.stderr, "##### EvaluateInteractionXML #####"
print >> sys.stderr, "Comparing input", inputCorpusFile, "to gold", goldCorpusFile
# Class sets are used to convert the types to ids that the evaluator can use
classSets = {}
if EvaluatorClass.type == "binary":
classSets["entity"] = IdSet(idDict={
"True": 1,
"False": -1
},
locked=True)
classSets["interaction"] = IdSet(idDict={
"True": 1,
"False": -1
},
locked=True)
negativeClassId = -1
elif EvaluatorClass.type == "multiclass":
classSets["entity"] = IdSet(idDict={"neg": 1}, locked=False)
classSets["interaction"] = IdSet(idDict={"neg": 1}, locked=False)
negativeClassId = 1
else:
sys.exit("Unknown evaluator type")
# Load corpus and make sentence graphs
goldCorpusElements = None
if goldCorpusFile != None:
goldCorpusElements = CorpusElements.loadCorpus(
goldCorpusFile, parse, tokenization,
removeIntersentenceInteractions)
predictedCorpusElements = CorpusElements.loadCorpus(
inputCorpusFile, parse, tokenization, removeIntersentenceInteractions)
# Compare the corpora and print results on screen
return processCorpora(EvaluatorClass, predictedCorpusElements,
goldCorpusElements, target, classSets,
negativeClassId, entityMatchFunction)
def fromXML(self, input, parse, tokenization=None):
self.names = {}
if type(input) == types.StringType:
corpus = CorpusElements.loadCorpus(input, parse, tokenization)
else:
corpus = input
for sentence in corpus.sentences:
tokenTuples = self.prepareTokens(sentence.tokens)
for entity in sentence.entities:
if entity.get("isName") == "True":
tokens = self.getTokens(entity, tokenTuples)
assert len(tokens) > 0
self.addName(tokens)
self.addName(["".join(tokens)])
def gifxmlToGenia(input,
output,
task=1,
outputIsA2File=False,
submission=False,
verbose=True,
strengths=False):
assert (task == 1 or task == 2 or task == 3)
outputTarFilename = None
# Make or clear output directory
if verbose: print >> sys.stderr, "Writing shared task files",
if not outputIsA2File:
if output.find("tar.gz") != -1:
outputTarFilename = output
output = "temp-genia-format"
if os.path.exists(output):
if verbose: print >> sys.stderr, "over existing directory", output
shutil.rmtree(output)
else:
if verbose: print >> sys.stderr, "to directory", output
os.mkdir(output)
# Convert the gifxml to the genia format files
inputCorpus = CorpusElements.loadCorpus(
input, removeIntersentenceInteractions=False)
processCorpus(inputCorpus,
output,
task,
outputIsA2File,
verbose=verbose,
strengths=strengths)
if submission:
if not outputIsA2File:
makeSubmissionFile(options.output,
output.split("/")[-1] + ".tar.gz")
else:
print >> sys.stderr, "Warning: Single a2-file output, no submission package created"
if outputTarFilename != None:
print >> sys.stderr, "Compressing output to", outputTarFilename
outputTarFile = tarfile.open(outputTarFilename, "w:gz")
allFiles = os.listdir(output)
tempCwd = os.getcwd()
os.chdir(output)
for file in allFiles:
outputTarFile.add(file)
os.chdir(tempCwd)
outputTarFile.close
def fromXML(self, input, parse, tokenization=None):
self.names = {}
if type(input) == types.StringType:
corpus = CorpusElements.loadCorpus(input, parse, tokenization)
else:
corpus = input
for sentence in corpus.sentences:
tokenTuples = self.prepareTokens(sentence.tokens)
for entity in sentence.entities:
if entity.get("isName") == "True":
tokens = self.getTokens(entity, tokenTuples)
assert len(tokens) > 0
self.addName(tokens)
self.addName(["".join(tokens)])
def run(EvaluatorClass, inputCorpusFile, goldCorpusFile, parse, tokenization=None, target="both", entityMatchFunction=compareEntitiesSimple, removeIntersentenceInteractions=False):
print >> sys.stderr, "##### EvaluateInteractionXML #####"
print >> sys.stderr, "Comparing input", inputCorpusFile, "to gold", goldCorpusFile
# Class sets are used to convert the types to ids that the evaluator can use
classSets = {}
if EvaluatorClass.type == "binary":
classSets["entity"] = IdSet(idDict={"True":1,"False":-1}, locked=True)
classSets["interaction"] = IdSet(idDict={"True":1,"False":-1}, locked=True)
negativeClassId = -1
elif EvaluatorClass.type == "multiclass":
classSets["entity"] = IdSet(idDict={"neg":1}, locked=False)
classSets["interaction"] = IdSet(idDict={"neg":1}, locked=False)
negativeClassId = 1
else:
sys.exit("Unknown evaluator type")
# Load corpus and make sentence graphs
goldCorpusElements = None
if goldCorpusFile != None:
goldCorpusElements = CorpusElements.loadCorpus(goldCorpusFile, parse, tokenization, removeIntersentenceInteractions)
predictedCorpusElements = CorpusElements.loadCorpus(inputCorpusFile, parse, tokenization, removeIntersentenceInteractions)
# Compare the corpora and print results on screen
return processCorpora(EvaluatorClass, predictedCorpusElements, goldCorpusElements, target, classSets, negativeClassId, entityMatchFunction)
def gifxmlToGenia(input, output, task=1, outputIsA2File=False, submission=False, verbose=True, strengths=False):
assert(task == 1 or task == 2 or task == 3)
outputTarFilename = None
# Make or clear output directory
if verbose: print >> sys.stderr, "Writing shared task files",
if not outputIsA2File:
if output.find("tar.gz") != -1:
outputTarFilename = output
output = "temp-genia-format"
if os.path.exists(output):
if verbose: print >> sys.stderr, "over existing directory", output
shutil.rmtree(output)
else:
if verbose: print >> sys.stderr, "to directory", output
os.mkdir(output)
# Convert the gifxml to the genia format files
inputCorpus = CorpusElements.loadCorpus(input, removeIntersentenceInteractions=False)
processCorpus(inputCorpus, output, task, outputIsA2File, verbose=verbose, strengths=strengths)
if submission:
if not outputIsA2File:
makeSubmissionFile(options.output, output.split("/")[-1] + ".tar.gz")
else:
print >> sys.stderr, "Warning: Single a2-file output, no submission package created"
if outputTarFilename != None:
print >> sys.stderr, "Compressing output to", outputTarFilename
outputTarFile = tarfile.open(outputTarFilename, "w:gz")
allFiles = os.listdir(output)
tempCwd = os.getcwd()
os.chdir(output)
for file in allFiles:
outputTarFile.add(file)
os.chdir(tempCwd)
outputTarFile.close
metavar="FILE")
optparser.add_option("-o",
"--output",
default=None,
dest="output",
help="Output directory")
optparser.add_option("-f",
"--folds",
type="int",
default=10,
dest="folds",
help="X-fold cross validation")
(options, args) = optparser.parse_args()
# Load corpus and make sentence graphs
corpusElements = CorpusElements.loadCorpus(options.input)
outputTrees = []
for i in range(options.folds):
newRoot = ET.Element("corpus")
for key in corpusElements.rootElement.attrib.keys():
newRoot.attrib[key] = corpusElements.rootElement.attrib[key]
outputTrees.append(newRoot)
print >> sys.stderr, "Reading document ids"
documentIds = []
for document in corpusElements.documents:
docId = document.attrib["id"]
assert (not docId in documentIds)
documentIds.append(docId)
