def testAnnotationsWriter(self):
annotations = yamlReader.load("data/forklift_open_ended.yaml")
print annotations
print annotations[0]
entireText, yamlData = esdcIo.toYaml(annotations[0].esdcs)
self.assertEqual(entireText, "Forklift stop.")
self.assertEqual(yamlData, [{'EVENT': {'r': 'stop', 'f': 'Forklift'}}])
entireText, yamlData = esdcIo.toYaml(annotations[1].esdcs)
self.assertEqual(entireText, "to the truck")
self.assertEqual(yamlData, [{'PATH': {'r': 'to', 'l': 'the truck'}}])
entireText, yamlData = esdcIo.toYaml(annotations[2].esdcs)
self.assertEqual(entireText, "Go between the truck and the pallet.")
self.assertEqual(yamlData, [{'EVENT': {'r': 'Go',
'l': {'PATH':{'r':'between',
'l': [{'OBJECT':{'f':'the truck'}},
{'OBJECT':{'f':'the pallet'}}]}}}}])
for annotatedSentence in annotations:
entireText, yamlData = esdcIo.toYaml(annotatedSentence.esdcs)
try:
rereadAnnotations = esdcIo.fromYaml(entireText, yamlData)
self.assertEqual(rereadAnnotations, annotatedSentence.esdcs)
except:
print "starting esdcs", [e.asPrettyMap() for e in annotatedSentence.esdcs]
print "text", entireText
print "data", yamlData
raise
def editDistance(esdc1, esdc2):
yaml1 = esdcIo.toYaml(esdc1)
yaml2 = esdcIo.toYaml(esdc2)
str1 = yaml.dump(yaml1)
str2 = yaml.dump(yaml2)
score = edit_distance(str1, str2)
score = float(score) / max([len(str1), len(str2)])
return 1 - score
def toYaml(self):
self.checkRep()
yamlData = {}
yamlData['assignmentId'] = self.assignmentId
yamlData['command'] = esdcIo.toYaml(self.esdcs)
if self.agent != None:
yamlData['agent'] = self.agent.toYaml()
if self.context != None:
yamlData['context'] = self.context.toYaml()
try:
g = [
groundings.toYaml(self.esdcToGroundings[esdc])
for esdc in self.flattenedEsdcs
]
except KeyError:
print esdc
raise
if any(len(x) != 0 for x in g):
yamlData['objectGroundings'] = g
yamlData['groundingIsCorrect'] = [
self.isGroundingCorrect(e) for e in self.flattenedEsdcs
]
yamlData['source'] = [self.getSource(e) for e in self.flattenedEsdcs]
return yamlData
def testPickUpCokeCan(self):
esdcs = self.extractor.extractEsdcsFromSentence(
"""Pick up the Coke can""")
self.aeq(
toYaml(esdcs)[1], [{
'EVENT': {
'r': 'Pick up',
'l': 'the Coke can'
}
}])
def testNestedRepeatedStrings(self):
from esdcs.dataStructures import ExtendedSdc, ExtendedSdcGroup
from standoff import TextStandoff
txt = "Move to the right side of the trailer of the trailer on the right and wait."
esdcs = [ExtendedSdc('EVENT', r=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (0, 4))],l2=[],l=[ExtendedSdc('PATH', r=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (5, 7))],l2=[],l=[ExtendedSdc('OBJECT', r=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (23, 25))],l2=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (41, 44)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (45, 52))])],l=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (26, 29)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (30, 37))])],f=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (8, 11)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (12, 17)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (18, 22))])])],f=[])],f=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[], entireText='Move to the right side of the trailer of the trailer on the right and wait.')]), ExtendedSdc('OBJECT', r=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (53, 55))],l2=[],l=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (56, 59)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (60, 65))])],f=[ExtendedSdc('OBJECT', r=[],l2=[],l=[],f=[TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (41, 44)), TextStandoff("Move to the right side of the trailer of the trailer on the right and wait.", (45, 52))])])]
entireText, yamlData = esdcIo.toYaml(ExtendedSdcGroup(esdcs))
rereadAnnotations = esdcIo.fromYaml(entireText, yamlData)
try:
self.assertEqual(list(rereadAnnotations), esdcs)
except:
print "start with", [e.asPrettyMap() for e in esdcs]
print "ended with", [e.asPrettyMap() for e in rereadAnnotations]
raise
def main():
parser = OptionParser()
parser.add_option("--source_corpus",
dest="source_corpus",
help="The file or directory to read from")
parser.add_option("--dest_corpus",
dest="dest_corpus",
help="The yaml file to which data will be written")
(options, args) = parser.parse_args()
corpus = readCorpus.Corpus(options.source_corpus)
extractor = Extractor()
#oldCorpus = annotationIo.load("%s/dataAnnotation/data/forkliftMturkEsdcs.stefie10.groundings.withPaths.yaml" % os.environ['FORKLIFT_HOME'])
#oldAssignmentIds = set(a.assignmentId for a in oldCorpus)
#print len(oldAssignmentIds), "old ids"
yamlData = []
print len(corpus.assignments), "assignments"
filteredAnnotations = [
x for x in corpus.assignments if (x.scenario.platform == "forklift"
and x.scenario.isSimulated == True)
]
print len(filteredAnnotations), "filtered annotations"
output_annotations = []
for i, a in enumerate(filteredAnnotations):
esdcs = extractor.extractEsdcs(a.command)
assignmentId = a.assignmentId
agent = a.agent
context = a.context
esdcSource = [assignmentId for esdc in esdcs.flattenedEsdcs]
output_annotations.append(
Annotation(assignmentId=assignmentId,
esdcs=esdcs,
objectGroundings=None,
groundingIsCorrect=None,
agent=agent,
context=context,
esdcSource=esdcSource))
esdcYaml = toYaml(esdcs)
print "******"
print a, esdcYaml
print esdcs
yamlData.append({"command": esdcYaml, "assignmentId": a.assignmentId})
print "dumped", i, "commands."
yaml.dump(yamlData, open(options.dest_corpus, "w"))
def ain(self, esdc_groups, correct_yaml):
current_metadata = None
parse_i = 0
lst = [toYaml(esdcs)[1] for esdcs in esdc_groups]
for i, l in enumerate(lst):
esdc = esdc_groups[i]
if esdc.metadata != current_metadata:
print "****************************** dependency", parse_i,
current_metadata = esdc.metadata
print current_metadata[0]
parse_i += 1
if correct_yaml == l:
print "correct"
else:
print "incorrect"
print l
if not correct_yaml in lst:
self.fail("Correct esdc not in results.")
def testCrash(self):
esdcs = self.extractor.extractEsdcs(
"""The trailer is to your right, and in at the end of the row of pallets."""
)
self.aeq(
toYaml(esdcs)[1], [{
'EVENT': {
'r': 'is',
'l2': {
'PLACE': {
'r': 'in',
'l': 'your right'
}
},
'l': {
'PATH': {
'r': 'to',
'l': 'your right'
}
},
'f': 'The trailer'
}
}, {
'PLACE': {
'r': 'at',
'l': {
'OBJECT': {
'r': [['of', [48, 50]]],
'l2': 'pallets',
'l': 'the row',
'f': 'the end'
}
}
}
}])
def followCommand(self,
command=None,
state=None,
esdcs=None,
first_esdc_only=False,
verbose=True,
input_gggs=None):
if command != None:
self.updateCommand(command)
if state != None:
self.state = state
command = str(self.commandEdit.toPlainText())
if input_gggs != None:
self.esdcs = input_gggs[0].esdcs
elif esdcs != None and command == self.command:
self.esdcs = esdcs
else:
class AnnotationStub:
def __init__(self, command):
self.entireText = command
self.esdcs = self.extractor(AnnotationStub(command))
if len(self.esdcs) == 0:
return [], []
if first_esdc_only:
self.esdcs = ExtendedSdcGroup([self.esdcs[0]])
self.flattenedEsdcs = self.esdcs.flattenedEsdcs
print "esdcs", esdcIo.toYaml(self.esdcs)
#print "flattened esdcs", self.flattenedEsdcs
self.esdcModel.setData(self.esdcs)
self.esdcView.expandAll()
#convert to a list of plans
if input_gggs != None:
gggs = input_gggs
elif self.gggs != None:
gggs = self.gggs
else:
if self.merging_mode == "merge_coreferences":
from coreference.bag_of_words_resolver import BagOfWordsResolver
from coreference.merge_coreferences import merge_coreferences
resolver = BagOfWordsResolver(
"%s/tools/coreference/models/coref_1.5.pck" % SLU_HOME)
gggs = merge_coreferences(self.esdcs, resolver)
if len(gggs) > 1:
gggs = [gggs[0]]
elif self.merging_mode == "merge_events":
from coreference.event_resolver import EventResolver
from coreference.merge_coreferences import merge_coreferences
resolver = EventResolver()
gggs = merge_coreferences(self.esdcs, resolver)
gggs = gggs[0:1]
for i, ggg in enumerate(gggs):
#ggg.to_latex("ggg_%d.pdf" % i)
pass
# export parses, when saving ground truth parses from asr evaluation run
# -- stefie10 7/19/2012
#from esdcs.esdcIo import toYaml
#import yaml
#import os
#with open("esdcs_%d.txt" % os.getpid(), "a") as f:
# f.write(yaml.dump([toYaml(self.esdcs)]))
if len(self.esdcs) != 1:
raise ValueError("ESDCs didn't parse right: " +
` self.esdcs `)
# print "merging events", self.esdcs.entireText
# from coreference.merge_coreferences import merge_toplevel_events
# #resolver = EventResolver()
# #gggs = merge_coreferences(self.esdcs, resolver,
# # verbose=False)
# gggs = gggs_from_esdc_group(self.esdcs)
# gggs = [merge_toplevel_events(gggs)]
assert len(gggs) == 1, (len(gggs), self.esdcs.text)
elif self.merging_mode == "merge_none":
gggs = gggs_from_esdc_group(self.esdcs)
else:
raise ValueError("Bad merging mode: " + ` self.merging_mode `)
if gggs[0].context == None:
gggs[0].context = self.state.to_context()
assert gggs[0].context.agent != None
def run():
try:
assert gggs[0].context.agent != None
self.plansList = self.taskPlanner.find_plan(
self.state,
gggs,
beam_width=self.beamWidthBox.value(),
beam_width_sequence=self.seqBeamWidthBox.value(),
search_depth_event=self.searchDepthBox.value(),
beam_width_event=self.beamWidthEventBox.value(),
save_state_tree=self.save_state_tree,
allow_null_action=self.allow_null_action)
except:
print "excption on gggs"
#for i, ggg in enumerate(gggs):
# ggg.to_latex("ggg_%d.pdf" % i)
raise
start = time.time()
cProfile.runctx("run()", globals(), locals(), "out.prof")
end = time.time()
if verbose:
print "Cost Function Browser took", (end - start), "seconds."
plansList = self.plansList
if len(plansList) == 0:
return [], []
else:
self.plans = plansModel.Plan.from_inference_result(
self.taskPlanner, plansList)
self.plansModel.setData(self.plans)
self.nodeFeatureWeights.load(self.taskPlanner.cf_obj,
gggs,
self.plans,
context=self.state.to_context())
self.plansView.selectRow(0)
self.gggWindow.load(plansList[0][2],
groundingSpace=self.state.objects)
return self.esdcs, self.plans
def testGoLeft(self):
command = "go left"
esdc_group = self.extractor.extractEsdcs(command)
self.aeq(toYaml(esdc_group)[1], [{'EVENT': {'r': 'go left'}}])
def testOrdering(self):
esdcs = self.extractor.extractEsdcs(
"""Come out of the parking lot and turn left. You''ll come to Titus Avenue. Take
a right, and then a quick left onto Hudson. Wegmans will be on the right."""
)
print esdcs.__class__
self.aeq(
toYaml(esdcs)[1], [{
'EVENT': {
'r': 'Come',
'l': {
'PATH': {
'r': 'out of',
'l': 'the parking lot'
}
}
}
}, {
'EVENT': {
'r': 'turn left'
}
}, {
'EVENT': {
'r': [['You', [43, 46]], ['come', [51, 55]]],
'l': {
'PATH': {
'r': [['to', [56, 58]]],
'l': 'Titus Avenue'
}
},
'f': [['ll', [48, 50]]]
}
}, {
'EVENT': {
'r': 'Take',
'l2': {
'PATH': {
'r': 'onto',
'l': 'Hudson'
}
},
'l': [['a', [78, 79]], ['right', [80, 85]],
['then', [91, 95]]]
}
}, {
'EVENT': {
'r': [['a', [78, 79]], ['right', [80, 85]],
['then', [91, 95]]]
}
}, {
'EVENT': {
'r': 'a quick left'
}
}, {
'EVENT': {
'r': 'will be',
'l': {
'PLACE': {
'r': [['on', [138, 140]]],
'l': 'the right'
}
},
'f': 'Wegmans'
}
}])