def getFeatureVector(poem):
# Takes in an author vector and a poem string. extracts the feature vector
# for this poem. This will mainly take the L1 norm between the poem characteristics
# and the author characteristics
phi = {}
poemVector = poemFeatures.poemCharacter(("",poem)) # Don't need author name or word pairs
phi['numLines'] = abs(poemVector['numLines'])
phi['avgWordLength'] = abs(poemVector['avgWordLength'])
phi['avgLineLength'] = abs(poemVector['avgLineLength'])
phi['rhymePercentAA'] = abs(poemVector['rhymePercentAA'])
phi['rhymePercentABA'] = abs(poemVector['rhymePercentABA'])
return phi
def styleTrainer():
testVectors = []
authorVectors = {}
data = readUtil.getTrainingData()
trainingSet = []
testingSet = []
# go through each author in data
for author in data.keys():
# initialize values for that author... could be simplified by making a vector class...
halfPoems = len(data[author])/2
authorVectors[author] = {}
authorVectors[author]['author'] = author
authorVectors[author]['numLines'] = 0.0
authorVectors[author]['avgWordLength'] = 0.0
authorVectors[author]['avgLineLength'] = 0.0
authorVectors[author]['rhymePercentAA'] = 0.0
authorVectors[author]['rhymePercentABA'] = 0.0
authorVectors[author]['wordPairs'] = collections.Counter()
authorVectors[author]['wordsPerLine'] = collections.Counter()
authorVectors[author]['linesPerPoem'] = collections.Counter()
authorVectors[author]['typeTokenCount'] = collections.Counter()
authorVectors[author]['wordDomain'] = collections.Counter()
authorVectors[author]['poemStart'] = collections.Counter()
# go through all of poems
for index in range(len(data[author])):
# get stats on the poem
poemVector = poemFeatures.poemCharacter(data[author][index])
# first half is for training
if index < halfPoems:
trainingSet.append(data[author][index])
# add those stats to the author
authorVectors[author]['numLines'] += poemVector['numLines']
authorVectors[author]['avgWordLength'] += poemVector['avgWordLength']
authorVectors[author]['avgLineLength'] += poemVector['avgLineLength']
authorVectors[author]['rhymePercentAA'] += poemVector['rhymePercentAA']
authorVectors[author]['rhymePercentABA'] += poemVector['rhymePercentABA']
authorVectors[author]['wordPairs'].update(poemVector['wordPairs'])
authorVectors[author]['wordsPerLine'].update(poemVector['wordsPerLine'])
authorVectors[author]['linesPerPoem'][poemVector['numLines']] += 1
authorVectors[author]['typeTokenCount'][poemVector['typeTokenCount']] += 1
authorVectors[author]['wordDomain'].update(poemVector['wordDomain'])
authorVectors[author]['poemStart'].update(poemVector['poemStart'])
# second half goes into testing data
else:
testVectors.append(poemVector)
testingSet.append(data[author][index])
# get averages
authorVectors[author]['numLines'] = authorVectors[author]['numLines']/halfPoems
authorVectors[author]['avgWordLength'] = authorVectors[author]['avgWordLength']/halfPoems
authorVectors[author]['avgLineLength'] = authorVectors[author]['avgLineLength']/halfPoems
authorVectors[author]['rhymePercentAA'] = authorVectors[author]['rhymePercentAA']/halfPoems
authorVectors[author]['rhymePercentABA'] = authorVectors[author]['rhymePercentABA']/halfPoems
return authorVectors, testVectors, trainingSet, testingSet
