def evalRelationTypes(fname, fGoldStd, methodLabel, method ):
""" evaluates the given ontology and writes the results into a file
@param[in] fname file name of the ontology to evaluate
@param[in] fGoldStd file name of the gold standard ontology
@param[in] methodLabel label of the method used in the evaluation
@param[in] method method used in the evaluator
"""
goldStd = _readOntology( fGoldStd )
ontology = _readOntology( fname )
goldStdConcepts = set(map(str, extractRelationSet(goldStd)))
ontologyConcepts = set(map(str, extractRelationSet(ontology)))
log.info("Comparing the relation set %s to the gold standard %s." % (ontologyConcepts, goldStdConcepts))
res = [ 1 ]
for scoringMethod in (EqualRel, EqualGroup, SimilarGroup):
__cache__ = DiskCache(".diskCache-%s-%s" % (scoringMethod.__name__, os.path.basename(fGoldStd)) )
c = ConceptScoring(ontologyConcepts, goldStdConcepts, scoringMethod, '|')
key = "%s, %s |" % (ontologyConcepts, goldStdConcepts)
score = __cache__.fetchObjectId(key, c.score)
res.append(score)
# compute precision and recall
p = float(score) / len(ontologyConcepts)
r = float(score) / len(goldStdConcepts)
if p==0. and r== 0.:
res.append(0.)
else:
res.append( metrics.fMeasure(p,r) )
#print ">>>", len(goldStdConcepts), len(ontologyConcepts), "***", res
return res
def testDirectCall(self):
''' tests directly calling the cache object using __call__ '''
CACHE_DIR = get_cache_dir(4)
cached_str = DiskCache(CACHE_DIR, fn=str)
assert cached_str(7) == "7"
assert cached_str.getKey(7) in cached_str
def testDirectCall(self):
''' tests directly calling the cache object using __call__ '''
CACHE_DIR = get_cache_dir(4)
cached_str = DiskCache(CACHE_DIR, fn=str)
assert cached_str(7) == "7"
assert cached_str.getKey(7) in cached_str
def __init__(self, dataSource, cache=True):
""" @param[in] dataSource implementing the TagInfoService Interface """
assert isinstance(dataSource, TagInfoService)
self.dataSource = dataSource
if cache == True:
diskCache = DiskCache("./.coherence-tagcount-cache", 2)
self.getTagCount = lambda tt: diskCache.fetchObjectId(self.dataSource.__class__.__name__ + str(tt),
self.dataSource.getTagInfo, tt)
else:
self.getTagCount = self.dataSource.getTagInfo
def __init__(self, dataSource, cache=True):
""" @param[in] dataSource implementing the TagInfoService Interface """
assert isinstance(dataSource, TagInfoService)
self.dataSource = dataSource
if cache == True:
diskCache = DiskCache("./.coherence-tagcount-cache", 2)
self.getTagCount = lambda tt: diskCache.fetchObjectId(self.dataSource.__class__.__name__ + str(tt),
self.dataSource.getTagInfo, tt)
else:
self.getTagCount = self.dataSource.getTagInfo
def testObjectKeyGeneration(self):
''' ensures that the diskcache object's location does not change '''
CACHE_DIR = get_cache_dir(3)
d = DiskCache(CACHE_DIR)
getCacheLocation = lambda x: join(CACHE_DIR, Cache.getObjectId(x))
d.fetchObjectId(1, str, 1)
assert exists( getCacheLocation(1) )
d.fetch(str, 2)
assert exists( getCacheLocation( ((2,), ()) ))
def getPostHashfile(self, cmd):
''' returns an identifier representing the object which is compatible
to the identifiers returned by the eWRT.util.cache.* classes. '''
args = (
tuple(cmd[1:]), ()
) # required to produce the same hash as DiskCache's fetch method
return self._get_fname(DiskCache.getObjectId(args))
class WebDocumentTerm(TermReference):
""" @class WebDocumentTerm
Similarity metric based on the similarity of the documents retrieved
with a web search.
"""
yahoo = Yahoo()
__cache__ = DiskCache(".diskCache-WebDocumentTerm-conceptCache",
cache_nesting_level=2)
@staticmethod
def _getConceptWebDocuments(concept):
""" returns web documents describing the given concept
@param[in] concept concept used to describe the text
"""
searchTerms = (concept.name, ) + tuple(
concept.context_terms)[:CONTEXT_TERM_COUNT]
log.debug("Searching for %s" % str(searchTerms))
yq = Yahoo.getSearchResults( \
WebDocumentTerm.yahoo.query( searchTerms, \
count=WEB_DOCUMENT_COUNT, \
queryParams={'view':'keyterms', 'abstract': 'long', 'type':'html,text'}) )
p = Pool(WEB_DOCUMENT_COUNT)
text = "\n".join(p.map(p_getWebDocumentText, yq))
return cleanup(text)
@staticmethod
def _getConceptWebDocumentsVector(concept):
return VectorSpaceModel(
WebDocumentTerm._getConceptWebDocuments(concept).split())
@staticmethod
@DiskCached(".diskCache-WebDocumentTerm-or")
def _or(c1, c2):
""" Compares two concepts and returns their similarity score
@param[in] c1 the first OntologyConcept
@param[in] c2 the second OntologyConcept
@returns the similarity betwen c1 and c2
"""
c1Text = WebDocumentTerm.__cache__.fetchObjectId(
c1, WebDocumentTerm._getConceptWebDocumentsVector, c1)
c2Text = WebDocumentTerm.__cache__.fetchObjectId(
c2, WebDocumentTerm._getConceptWebDocumentsVector, c2)
# similarity for concepts with no matches
if len(c1Text.v) == 0 or len(c2Text.v) == 0:
if len(c1Text.v) == 0:
log.warn("No web pages found for '%s'" % c1)
if len(c2Text.v) == 0:
log.warn("No web pages found for '%s'" % c2)
return 0.
return c1Text * c2Text
def __or__(self, o):
return self._or(self.e, o.e)
def evalOntology( fname, fGoldStd, methodLabel, method ):
""" evaluates the given ontology and writes the results into a file
@param[in] fname file name of the ontology to evaluate
@param[in] fGoldStd file name of the gold standard ontology
@param[in] methodLabel label of the method used in the evaluation
@param[in] method method used in the evaluator
"""
goldStd = _readOntology( fGoldStd )
ontology = _readOntology( fname )
goldStdConcepts = OntologyConcept.sequenceToOntologyConceptList(extractConceptSet(goldStd))
ontologyConcepts = OntologyConcept.sequenceToOntologyConceptList(extractConceptSet(ontology))
log.info("Comparing the ontology concepts %s to the gold standard %s." % (ontologyConcepts, goldStdConcepts))
res = [ conceptTermCount( ontology ) ]
for scoringMethod in (EqualTerm, StringEditTerm, PhoneticTerm, WordNetTerm, WikipediaTerm, WebDocumentTerm, GoogleDistanceTerm, OntologyTerm, ):
__cache__ = DiskCache(".diskCache-%s-%s" % (scoringMethod.__name__, os.path.basename(fGoldStd)) )
# Methods using neighbor concepts
if scoringMethod in (WebDocumentTerm, ):
goldNeighborConcepts = OntologyConcept.statementsToDirectNeighborOntologyConceptList( extractSPO(goldStd) )
ontoNeighborConcepts = OntologyConcept.statementsToDirectNeighborOntologyConceptList( extractSPO(ontology) )
c = ConceptScoring(ontoNeighborConcepts, goldNeighborConcepts, scoringMethod, '|', poolSize=1)
key = "%s, %s |" % (ontoNeighborConcepts, goldNeighborConcepts)
# methods using all concepts
else:
ps = 1 if scoringMethod == OntologyTerm else 4
c = ConceptScoring(ontologyConcepts, goldStdConcepts, scoringMethod, '|', poolSize=ps)
key = "%s, %s |" % (ontologyConcepts, goldStdConcepts)
score = __cache__.fetchObjectId(key, c.score)
print scoringMethod, score
res.append(score)
# compute precision and recall
p = float(score) / len(ontologyConcepts)
r = float(score) / len(goldStdConcepts)
if p==0. and r== 0.:
res.append(0.)
else:
res.append( metrics.fMeasure(p,r) )
return res
def __init__(self, submitter, api_key=OPENCALAIS_KEY, allow_distro="false", allow_search="false", cache_dir=OPENCALAIS_CACHE_DIR):
"""
Creates a new handler for communicating with OpenCalais.
The parameter 'submitter' must contain a string, identifying your application.
'api_key' must contain a string with your OpenCalais API key (get it here: http://developer.opencalais.com/apps/register).
The optional parameter 'allow_distro', if set to 'true' gives OpenCalais permission to distribute the metadata extracted from your submissions. The default value for 'allow_distro' is 'false'.
The optional parameter 'allow_search', if set to 'true' tells OpenCalais that future searches can be performed on the extracted metadata. The default value for 'allow_search' is 'false'.
"""
assert(api_key)
self.submitter = submitter
self.allow_distro = "false"
self.allow_search = "false"
self.api_key = api_key
if cache_dir:
self.cache = DiskCache(cache_dir, cache_nesting_level=2, cache_file_suffix=".xml")
class Calais:
submitter = USER_AGENT % "Calais"
allow_distro = "false"
allow_search = "false"
api_key = ""
def __init__(self, submitter, api_key=OPENCALAIS_KEY, allow_distro="false", allow_search="false", cache_dir=OPENCALAIS_CACHE_DIR):
"""
Creates a new handler for communicating with OpenCalais.
The parameter 'submitter' must contain a string, identifying your application.
'api_key' must contain a string with your OpenCalais API key (get it here: http://developer.opencalais.com/apps/register).
The optional parameter 'allow_distro', if set to 'true' gives OpenCalais permission to distribute the metadata extracted from your submissions. The default value for 'allow_distro' is 'false'.
The optional parameter 'allow_search', if set to 'true' tells OpenCalais that future searches can be performed on the extracted metadata. The default value for 'allow_search' is 'false'.
"""
assert(api_key)
self.submitter = submitter
self.allow_distro = "false"
self.allow_search = "false"
self.api_key = api_key
if cache_dir:
self.cache = DiskCache(cache_dir, cache_nesting_level=2, cache_file_suffix=".xml")
@staticmethod
def random_id(self):
"""
Creates a random 10-character ID for your submission.
"""
chars = str.letters + str.digits
return "".join( [ choice(chars) for i in xrange(10) ] )
@staticmethod
def content_id(text):
"""
Creates a SHA1 hash of the text of your submission.
"""
try:
import hashlib
h = hashlib.sha1()
except ImportError:
import sha
h = sha.new()
h.update(text)
return h.hexdigest()
def analyze(self, text, content_type="text/txt"):
""" Submits 'text' to OpenCalais for analysis and memorizes the extracted metadata.
Set the content-type to 'text/html' if you are submitting HTML data.
"""
externalID = self.content_id( text )
paramsXML = PARAMS_XML % (content_type, self.allow_distro, self.allow_search, externalID, self.submitter)
param = urlencode({'licenseID':self.api_key, 'content':text, 'paramsXML':paramsXML})
# do not fetch the data again, if a file exists in the cache
get_calais_data = lambda x: Retrieve(Calais.__name__).open(OPENCALAIS_URL, x).read()
if self.cache is None:
xml_data = self.unpack( get_calais_data( param ) )
else:
xml_data = self.unpack( self.cache.fetch( get_calais_data, param ) )
return self.parse( xml_data )
@staticmethod
def unpack(calais_data):
""" extracts calais' xml response from the data send by the calais
webservice
"""
dom = minidom.parseString(calais_data)
return """<?xml version="1.0" encoding="utf-8"?>\n""" \
+ dom.getElementsByTagName("string")[0].firstChild.data
@staticmethod
def cleanup_xml(xml_data):
""" removes comments from xml-data-streams provided by opencalais
@param[in] xml_data
@returns the xml data without any comments
"""
result = []
comment = False
while '<!--' in xml_data:
xml_data = re.sub('<!--[\s\S]*?-->', '', xml_data)
if not re.search('<!--', xml_data):
break
return xml_data
@staticmethod
def parse(xml_data):
""" parses opencalai's xml output and returns it's dictionary representation """
things = []
xml_data = Calais.cleanup_xml(xml_data)
# f= open("tmp","w"); f.write(xml_data.encode("utf8")); f.close()
dom = minidom.parseString( xml_data.encode("utf8" ))
for document in dom.getElementsByTagName("CalaisSimpleOutputFormat"):
for annotations in document.childNodes:
if not annotations.hasChildNodes():
continue
if annotations.nodeName == 'Topics':
annotations = annotations.firstChild
nodeName = annotations.nodeName
nodeAttr = dict(annotations.attributes.items())
nodeAttr.update( {'data': annotations.firstChild.data } )
things.append( {nodeName: nodeAttr } )
return things
def getPostHashfile(self, cmd):
''' returns an identifier representing the object which is compatible
to the identifiers returned by the eWRT.util.cache.* classes. '''
args = (tuple(
cmd[1:]), ()) # required to produce the same hash as DiskCache's fetch method
return self._get_fname(DiskCache.getObjectId(args))
def setUp(self):
self.diskCache = DiskCache(get_cache_dir(4))
class SkipTestDiskCached(TestCached):
@staticmethod
@DiskCached(get_cache_dir(1))
def add(a=1, b=2):
return a+b
@staticmethod
@DiskCached(get_cache_dir(2))
def sub(a, b):
return a-b
def setUp(self):
self.diskCache = DiskCache(get_cache_dir(4))
def tearDown(self):
''' remove the cache directories '''
for cacheDirNo in range(10):
if exists(get_cache_dir(cacheDirNo)):
rmtree(get_cache_dir(cacheDirNo))
def testObjectKeyGeneration(self):
''' ensures that the diskcache object's location does not change '''
CACHE_DIR = get_cache_dir(3)
d = DiskCache(CACHE_DIR)
getCacheLocation = lambda x: join(CACHE_DIR, Cache.getObjectId(x))
d.fetchObjectId(1, str, 1)
assert exists( getCacheLocation(1) )
d.fetch(str, 2)
assert exists( getCacheLocation( ((2,), ()) ))
def testContains(self):
''' verifies that 'key' in cache works '''
# diskcache
assert self.diskCache.fetchObjectId(1, str, 1 ) == "1"
assert 1 in self.diskCache
assert 2 not in self.diskCache
# diskcached
assert self.add(12,14) == 26
assert self.add.getKey(12,14) in self.add
assert 9 not in self.add
def testDelItem(self):
''' verifies that delitem works '''
# diskcache
assert self.diskCache.fetch(str, 2) == "2"
key = self.diskCache.getKey(2)
assert key in self.diskCache
del self.diskCache[key]
assert key not in self.diskCache
# diskcached
assert self.add(12,13) == 25
key = self.add.getKey(12, 13)
assert key == ((12, 13), ())
assert key in self.add
del self.add[key]
assert key not in self.add
def testDirectCall(self):
''' tests directly calling the cache object using __call__ '''
CACHE_DIR = get_cache_dir(4)
cached_str = DiskCache(CACHE_DIR, fn=str)
assert cached_str(7) == "7"
assert cached_str.getKey(7) in cached_str
def testIterableCache(self):
''' tests the iterable cache '''
CACHE_DIR = get_cache_dir(5)
i = IterableCache(CACHE_DIR)
getTestIterator = lambda x: range(x)
for iteratorSize in (4, 5, 6):
cachedIterator = i.fetch( getTestIterator, iteratorSize )
for x,y in zip(cachedIterator, getTestIterator(iteratorSize)):
assert x == y
@pytest.mark.slow
def testThreadSafety(self):
''' tests whether everything is thread safe '''
for a in range(1000):
c = DiskCache(get_cache_dir(6))
p = Pool(12)
p.map(f, 60*[c] )
p.map(g, 60*[c] )
p.close()
p.join()
def __init__(self, e):
""" @param[in] lm A list of TermReference metrics to use in the ontology metric """
TermReference.__init__(self, e)
self.metrics = dict([(m(e),
DiskCache(".diskCache-single-%s" % m.__name__))
for m in self.METRICS])