import numpy as np
import nltk # for pos tags
import features
import polarity
import ngramGenerator
import preprocessing
KERNEL_FUNCTION='linear'
C_PARAMETER=0.6
print "Initializing dictionnaries"
stopWords = preprocessing.getStopWordList('../resources/stopWords.txt')
slangs = preprocessing.loadSlangs('../resources/internetSlangs.txt')
afinn=polarity.loadAfinn('../resources/afinn.txt')
#sentiWordnet=polarity.loadSentiWordnet('../resources/sentiWordnetBig.csv')
emoticonDict=features.createEmoticonDictionary("../resources/emoticon.txt")
print "Bulding Bag of words ..."
positive=ngramGenerator.mostFreqList('../data/used/positive1.csv',3000)
negative=ngramGenerator.mostFreqList('../data/used/negative1.csv',3000)
neutral=ngramGenerator.mostFreqList('../data/used/neutral1.csv',3000)
for w in positive:
if w in negative+neutral :
positive.remove(w)
for w in negative:
if w in positive+neutral :
import features
import polarity
import ngramGenerator
import preprocessing
# User input for model parameters
N_NEIGHBORS = 10 # number of neighbors for KNN
KERNEL_FUNCTION = 'linear' # kernel function for SVM
C_PARAMETER = 0.2
UNIGRAM_SIZE = 3000
print "Initializing dictionnaries"
stopWords = preprocessing.getStopWordList('../resources/stopWords.txt')
slangs = preprocessing.loadSlangs('../resources/internetSlangs.txt')
afinn = polarity.loadAfinn('../resources/afinn.txt')
emoticonDict = features.createEmoticonDictionary("../resources/emoticon.txt")
print "Bulding unigram vector"
positive = ngramGenerator.mostFreqList('../data/used/positive1.csv',
UNIGRAM_SIZE) # add as needed
negative = ngramGenerator.mostFreqList('../data/used/negative1.csv',
UNIGRAM_SIZE)
neutral = ngramGenerator.mostFreqList('../data/used/neutral1.csv',
UNIGRAM_SIZE)
for w in positive:
if w in negative + neutral:
positive.remove(w)
for w in negative:
def predecir(tweet,
model): # prueba un tweet nuevo en base aun modelo ya creado
z = mapTweet(tweet, afinn, emoticonDict, positive, negative, neutral,
slangs)
z_scaled = scaler.transform([z])
z = normalizer.transform(z_scaled)
z = z[0].tolist()
return model.predict([z]).tolist()
# Preprocesamiento de los archivos
stopWords = preprocessing.getStopWordList(
abs_file_url('resources/stopWords.txt'))
slangs = preprocessing.loadSlangs(abs_file_url('resources/internetSlangs.txt'))
afinn = polarity.loadAfinn(abs_file_url('resources/afinn.txt'))
emoticonDict = features.createEmoticonDictionary(
abs_file_url('resources/emoticon.txt'))
# Se construye el vector con las palabras más frecuentes presentes en tweets positivos, negativos, y neutrales
positive = ngramGenerator.mostFreqList(abs_file_url('data/used/positive1.csv'),
3000)
negative = ngramGenerator.mostFreqList(abs_file_url('data/used/negative1.csv'),
3000)
neutral = ngramGenerator.mostFreqList(abs_file_url('data/used/neutral1.csv'),
3000)
# Normalizamos el tamaño de los unigramas, si es que son menores a 3000
min_len = min([len(positive), len(negative), len(neutral)])
positive = positive[0:min_len]