def compAssocForPairsInStemClass(stemClasses):
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
total = len(stemClasses)
for stem, classList in stemClasses.items():
#only consider stem classes with multiple members
if( len(classList) < 2 ):
continue
pairs = getPairs(classList)
print('stem:', stem)
print('\tstem class:', classList, '\n')
for i in range(len(pairs)):
'''
Compute a co-occurrence or association metric for each pair. This measures
how strong the association is between the words.
'''
dice = getAssociationForPair(vocabDict, pairs[i])
print('\tpair:', pairs[i])
print('\tdice:', dice)
print('\t', counter, 'of', total, '\n')
print()
counter += 1
def getStemsClassesSizeKPlus(k=2):
stemClasses = getDictFromFile('wiki-small-vocab-stem-classes.json')
chosenStemClasses = {}
for stem, classList in stemClasses.items():
if (len(classList) >= k):
chosenStemClasses[stem] = classList
diff = len(stemClasses) - len(chosenStemClasses)
print('old:', len(stemClasses))
print('new:', len(chosenStemClasses))
print('getStemsClassesSizeKPlus() - diff:', diff, '\n')
return chosenStemClasses
def getTopKFromDict(k):
print('\n'*2)
print('top', k)
if( k<1 ):
return
outlinksDict = getDictFromFile('./outlinksDict.json')
result = sorted(outlinksDict.items(), key=lambda x: x[1], reverse=True)
for i in range(len(result)):
print(i+1, result[i])
if( i == k-1 ):
break
def getKAlphabeticalWords(k=1000):
index = getDictFromFile('wiki-small-vocab.json')
sortedKeys = list(index.keys())
sortedKeys.sort()
counter = 0
for i in range(len(sortedKeys)):
if (isWord(sortedKeys[i]) == True):
counter += 1
print(sortedKeys[i])
if (counter == 1000):
break
def optimizeStemClass(oldStemClass, windowSize, threshold):
'''
Algorithm from: Search Engines Information Retrieval in Practice (page 191-192)
'''
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
total = len(oldStemClass)
for stem, classList in oldStemClass.items():
pairs = getPairs(classList)
'''
Construct a graph where the vertices represent words and the edges are between
words whose co-occurrence metric is above a threshold T.
'''
G = nx.Graph()
G.add_nodes_from(classList)
for i in range(len(pairs)):
'''
Compute a co-occurrence or association metric for each pair. This measures
how strong the association is between the words.
'''
dice = getAssociationForPair(vocabDict, pairs[i], windowSize)
if (dice >= threshold):
G.add_edge(pairs[i][0], pairs[i][1])
if (counter % 10 == 0):
print(counter, 'of', total, 'dice:', dice, '\n')
if (len(G.edges()) != 0):
print('Graph:')
print('nodes:', G.nodes())
print('edges:', G.edges())
conComp = list(nx.connected_component_subgraphs(G))
print()
print('New stem class for stem:', stem, ':')
for subgraph in conComp:
subgraph = subgraph.nodes()
if (len(subgraph) > 1):
print('\t', subgraph)
counter += 1
def getStemclasses():
stemClasses = {}
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
for voc, vocDict in vocabDict.items():
stem = PorterStemmer.useStemer(voc)
stemClasses.setdefault(stem, [])
stemClasses[stem].append(voc)
if (counter % 10000 == 0):
print('\t', counter, voc)
counter += 1
dumpJsonToFile('wiki-small-vocab-stem-classes.json', stemClasses, False)
def genCSVFile():
outfile = open('sorted-1-2-gram.org.csv', 'w')
outfile.write('Rank,Term,Freq,C\n')
gramsDict = getDictFromFile('1-2-gram.json')
sortedDict = sorted(gramsDict.items(), key=lambda x: x[1], reverse=True)
total = 770552
rank = 1
for tup in sortedDict:
term, freq = tup
c = (freq / total) * rank
outfile.write(
str(rank) + ', ' + term + ', ' + str(freq) + ', ' +
str(round(c, 5)) + '\n')
rank += 1
outfile.close()
def optimizeStemClass(oldStemClass, windowSize, threshold):
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
total = len(oldStemClass)
for stem, classList in oldStemClass.items():
pairs = getPairs(classList)
G = nx.Graph()
G.add_nodes_from(classList)
for i in range(len(pairs)):
dice = getAssociationForPair(vocabDict, pairs[i], windowSize)
if (dice >= threshold):
G.add_edge(pairs[i][0], pairs[i][1])
if (counter % 10 == 0):
print(counter, 'of', total, 'dice:', dice, '\n')
if (len(G.edges()) != 0):
print('Graph:')
print('nodes:', G.nodes())
print('edges:', G.edges())
conComp = list(nx.connected_component_subgraphs(G))
print()
print('New stem class for stem:', stem, ':')
for subgraph in conComp:
subgraph = subgraph.nodes()
if (len(subgraph) > 1):
print('\t', subgraph)
counter += 1
def compAssocForPairsInStemClassThreshold(stemClasses, threshold):
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
total = len(stemClasses)
for stem, classList in stemClasses.items():
if (len(classList) < 2):
continue
G = nx.Graph()
G.add_nodes_from(classList)
pairs = getPairs(classList)
stemDice = 0
for i in range(len(pairs)):
dice = getAssociationForPair(vocabDict, pairs[i])
if (dice >= threshold):
G.add_edge(pairs[i][0], pairs[i][1])
stemDice = dice
if (len(G.edges()) != 0):
conComp = list(nx.connected_component_subgraphs(G))
print('stem:', stem)
print('\tdice:', stemDice)
print('\told stem class:', classList, '\n')
print('\tNew stem class for stem:')
for subgraph in conComp:
subgraph = subgraph.nodes()
if (len(subgraph) > 1):
print('\t', subgraph)
print()
counter += 1
def compAssocForPairsInStemClass(stemClasses):
vocabDict = getDictFromFile('wiki-small-vocab.json')
counter = 0
total = len(stemClasses)
for stem, classList in stemClasses.items():
if (len(classList) < 2):
continue
pairs = getPairs(classList)
print('stem:', stem)
print('\tstem class:', classList, '\n')
for i in range(len(pairs)):
dice = getAssociationForPair(vocabDict, pairs[i])
print('\tpair:', pairs[i])
print('\tdice:', dice)
print('\t', counter, 'of', total, '\n')
print()
counter += 1
def getAssocMeasuresDocs(a, N, k=10):
vocabDict = getDictFromFile('wiki-small-vocab.json')
a = a.lower()
if (a not in vocabDict):
print('term:', a, 'not in vocab')
return
aFileSet = set(vocabDict[a]['f'])
vocabDict[a]['MIM'] = -1
vocabDict[a]['EMIM'] = -1
vocabDict[a]['CHI-SQUARE'] = -1
vocabDict[a]['DICE'] = -1
Na = len(aFileSet)
for b, bDict in vocabDict.items():
if (b == a):
continue
bFileSet = set(bDict['f'])
Nb = len(bFileSet)
intersect = aFileSet & bFileSet
MIM = -1
EMIM = -1
dice = -1
chiSquare = -1
Nab = len(intersect)
NaTimesNb = Na * Nb
if (Nab != 0):
MIM = Nab / (Na * Nb)
dice = Nab / (Na + Nb)
EMIM = Nab * math.log(N * MIM, 10)
if (NaTimesNb != 0):
numer = Nab - (NaTimesNb / N)
chiSquare = (numer * numer) / NaTimesNb
bDict['MIM'] = MIM
bDict['EMIM'] = EMIM
bDict['CHI-SQUARE'] = chiSquare
bDict['DICE'] = dice
for sortCriteria in ['MIM', 'EMIM', 'CHI-SQUARE', 'DICE']:
print()
sort = sorted(vocabDict.items(),
key=lambda x: x[1][sortCriteria],
reverse=True)
sort = sort[:k]
print(a, 'vs')
counter = 1
for termDict in sort:
term, termDict = termDict
print('\t', counter, 'term:', term, sortCriteria + ':',
termDict[sortCriteria])
counter += 1
def getAssocMeasuresWindow(a, N, filename, k=10):
prev = datetime.now()
vocabDict = getDictFromFile(filename)
a = a.lower()
if (a not in vocabDict):
print('term:', a, 'not in vocab')
return
transformDocToWindowOpt(vocabDict, a)
totalVocab = len(vocabDict)
pos = 0
vocabDict[a]['MIM'] = -1
vocabDict[a]['EMIM'] = -1
vocabDict[a]['CHI-SQUARE'] = -1
vocabDict[a]['DICE'] = -1
for b, bDict in vocabDict.items():
pos += 1
if (b == a):
continue
count = countTerms(vocabDict[a]['f']['windows'], a, b)
Na = count['left']
Nab = count['both']
transformDocToWindowOpt(vocabDict, b)
count = countTerms(vocabDict[b]['f']['windows'], b, a)
Nb = count['left']
MIM = -1
EMIM = -1
dice = -1
chiSquare = -1
if (pos % 100 == 0):
print(pos, 'of', totalVocab)
print('\tNa:', Na, a)
print('\tNb:', Nb, b)
print('\tNab:', Nab)
delta = datetime.now() - prev
print('\ttotal seconds:', delta.seconds)
NaTimesNb = Na * Nb
if (Nab != 0):
MIM = Nab / (Na * Nb)
dice = Nab / (Na + Nb)
EMIM = Nab * math.log(N * MIM, 10)
if (NaTimesNb != 0):
numer = Nab - (NaTimesNb / N)
chiSquare = (numer * numer) / NaTimesNb
bDict['MIM'] = MIM
bDict['EMIM'] = EMIM
bDict['CHI-SQUARE'] = chiSquare
bDict['DICE'] = dice
for sortCriteria in ['MIM', 'EMIM', 'CHI-SQUARE', 'DICE']:
print()
sort = sorted(vocabDict.items(),
key=lambda x: x[1][sortCriteria],
reverse=True)
sort = sort[:k]
print(a, 'vs')
for termDict in sort:
term, termDict = termDict
print('\tterm:', term, sortCriteria + ':', termDict[sortCriteria])
