def test_aggregated_sentiments():
sub_clf = classifier.get_optimal_subjectivity_classifier()
pol_clf = classifier.get_optimal_polarity_classifier()
tweets = utils.get_pickles(2)
sentimentvalues = utils.get_sentimentvalues(2)
sub_train_tweets, sub_train_targets, _, _, sub_train_sentiments, _ = utils.make_subjectivity_train_and_test_and_targets(
tweets, sentimentvalues, splitvalue=1.0
)
pol_train_tweets, pol_train_targets, _, _, pol_train_sentiments, _ = utils.make_polarity_train_and_test_and_targets(
tweets, sentimentvalues, splitvalue=1.0
)
sub_predictions = sub_clf.classify(sub_train_tweets, sub_train_sentiments)
pol_predictions = pol_clf.classify(pol_train_tweets, pol_train_sentiments)
print pol_train_targets, pol_predictions
days, targets, predicts, total_frequencies = utils.temporally_aggregate_subjectivity(
sub_train_tweets, sub_predictions, targets=sub_train_targets
)
data = {"Targets": [days, targets], "Predictions": [days, predicts], "Frequencies": [days, total_frequencies]}
plotting.plot_subjectivity_aggregates(data, "aggregated_subjectivity")
days, targets, predicts, frequencies = utils.temporally_aggregate_polarity(
pol_train_tweets, pol_predictions, targets=pol_train_targets
)
for i in range(len(days)):
targets[i] = targets[i] * 1.0 / frequencies[i]
predicts[i] = predicts[i] * 1.0 / frequencies[i]
frequencies[i] = frequencies[i] * 1.0 / total_frequencies[i]
data = {"Targets": [days, targets], "Predictions": [days, predicts], "Frequencies": [days, frequencies]}
plotting.plot_polarity_aggregates(data, "aggregated_polarity")
def perform_grid_search_on_featureset_SA_and_PA():
datasetnr = 3
tweets = utils.get_pickles(datasetnr)
sentimentvalues = feat_utils.get_sentiment_values(datasetnr)
tweets = preprocessing.remove_link_classes(tweets)
tweets = preprocessing.lower_case(tweets)
tweets = preprocessing.remove_specialchars_round2(tweets)
train_tweets, train_targets, test_tweets, test_targets, train_sentimentvalues, test_sentimentvalues = utils.make_subjectivity_train_and_test_and_targets(
tweets, sentimentvalues
)
clf = SVM(train_tweets, train_targets, None)
clf.set_feature_set("SA", None)
clf.grid_search_on_text_features(file_postfix="subjectivity")
clf = NB(train_tweets, train_targets, None)
clf.set_feature_set("SA", None)
clf.grid_search_on_text_features(file_postfix="subjectivity")
clf = ME(train_tweets, train_targets, None)
clf.set_feature_set("SA", None)
clf.grid_search_on_text_features(file_postfix="subjectivity")
train_tweets, train_targets, test_tweets, test_targets, train_sentimentvalues, test_sentimentvalues = utils.make_polarity_train_and_test_and_targets(
tweets, sentimentvalues
)
clf = SVM(train_tweets, train_targets, None)
clf.set_feature_set("PA", None)
clf.grid_search_on_text_features(file_postfix="polarity")
clf = NB(train_tweets, train_targets, None)
clf.set_feature_set("PA", None)
clf.grid_search_on_text_features(file_postfix="polarity")
clf = ME(train_tweets, train_targets, None)
clf.set_feature_set("PA", None)
clf.grid_search_on_text_features(file_postfix="polarity")
def google_lexicon_lookup():
"""
Fetches the tweets and performs lexicon translatino and lookup.
"""
tweets = utils.get_pickles(0)
words_with_values = lexicon.perform_google_sentiment_lexicon_lookup(tweets)
print "Storing..."
utils.store_sentimentvalues(words_with_values, "models/google_sentimentvalues_random_dataset")
tweets = utils.get_pickles(1)
words_with_values = lexicon.perform_google_sentiment_lexicon_lookup(tweets)
print "Storing..."
utils.store_sentimentvalues(words_with_values, "models/google_sentimentvalues_rosenborg_dataset")
tweets = utils.get_pickles(2)
words_with_values = lexicon.perform_google_sentiment_lexicon_lookup(tweets)
print "Storing..."
utils.store_sentimentvalues(words_with_values, "models/google_sentimentvalues_erna_dataset")
def train_and_store_results(self):
"""
Trains the given model on the dataset using the three different models, and different feature sets. Stores the results of the runs.
"""
dataset = "random_dataset"
tweets = utils.get_pickles(dataset)
self.model.set_feature_set('A')
self.model.train_on_feature_set()
def pos_analyze():
"""
Unpickles preprocessed tweets and performs pos-analysis of them. Then stores the stats in a diagram.
"""
tweets = utils.get_pickles(3)
subjectivity_data, polarity_data = pos_tag_analyze(tweets)
plotting.plot_pos_analysis(subjectivity_data, "sub_analysis")
plotting.plot_pos_analysis(polarity_data, "pos_analysis")
return True
def train_and_test_subjectivity_and_polarity():
datasetnr = 3
tweets = utils.get_pickles(datasetnr)
sentimentvalues = feat_utils.get_sentiment_values(datasetnr)
tweets = preprocessing.remove_link_classes(tweets)
tweets = preprocessing.lower_case(tweets)
tweets = preprocessing.remove_specialchars_round2(tweets)
# train_subjectivity_and_test_on_feature_set(tweets, 'SA', sentimentvalues)
train_subjectivity_and_test_on_feature_set(tweets, "SB", sentimentvalues)
train_subjectivity_and_test_on_feature_set(tweets, "SC", sentimentvalues)
def re_analyze():
"""
Unpickles preprocessed tweets and performs reanalyzis of these, then stores stats.
"""
labels = ["random",'"rosenborg"','"erna solberg"']
data = {}
worddata = {}
for i in xrange(3):
tweets = utils.get_pickles(i)
analyzer = Analyzer(utils.annotated_datasets[i], tweets)
avg_list,words_list= analyzer.analyze()
print avg_list
worddata[labels[i]] = words_list
data[labels[i]] = avg_list
plotting.average_wordclasses(worddata, "averages")
plotting.detailed_average_wordclasses(data, "averages2")
def get_optimal_polarity_classifier():
"""
Trains and returns the optimal polarity classifier.
"""
tweets = utils.get_pickles(3)
tweets, targets = utils.make_polarity_targets(tweets)
vect_options = {
'ngram_range': (1,1),
'max_df': 0.5
}
tfidf_options = {
'sublinear_tf': False,
'use_idf': True,
'smooth_idf': True,
}
clf = SVM(tweets, targets, vect_options, tfidf_options)
clf.set_feature_set('PC2', features.get_google_sentiment_values(3))
clf.train_on_feature_set()
return clf
def train_and_test_dataset_increase():
datasetnr = 3
tweets = utils.get_pickles(datasetnr)
sentimentvalues = feat_utils.get_sentiment_values(datasetnr)
tweets = preprocessing.remove_link_classes(tweets)
tweets = preprocessing.lower_case(tweets)
tweets = preprocessing.remove_specialchars_round2(tweets)
accuracy_data = {
"NB(SA)": [],
"NB(SB)": [],
"NB(SC)": [],
"SVM(SA)": [],
"SVM(SB)": [],
"SVM(SC)": [],
"MaxEnt(SA)": [],
"MaxEnt(SB)": [],
"MaxEnt(SC)": [],
"NB(PA)": [],
"NB(PB)": [],
"NB(PC)": [],
"SVM(PA)": [],
"SVM(PB)": [],
"SVM(PC)": [],
"MaxEnt(PA)": [],
"MaxEnt(PB)": [],
"MaxEnt(PC)": [],
}
f1_data = {
"NB(SA)": [],
"NB(SB)": [],
"NB(SC)": [],
"SVM(SA)": [],
"SVM(SB)": [],
"SVM(SC)": [],
"MaxEnt(SA)": [],
"MaxEnt(SB)": [],
"MaxEnt(SC)": [],
"NB(PA)": [],
"NB(PB)": [],
"NB(PC)": [],
"SVM(PA)": [],
"SVM(PB)": [],
"SVM(PC)": [],
"MaxEnt(PA)": [],
"MaxEnt(PB)": [],
"MaxEnt(PC)": [],
}
for i in range(5, 101, 5):
print "=============================DATAPOINT NR. ", i, "========================================"
data = train_subjectivity_and_test_on_feature_set(tweets, "SA", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(SA)"].append(data["Naive Bayes"][0])
f1_data["NB(SA)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(SA)"].append(data["SVM"][0])
f1_data["SVM(SA)"].append(data["SVM"][3])
accuracy_data["MaxEnt(SA)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(SA)"].append(data["Maximum Entropy"][3])
data = train_subjectivity_and_test_on_feature_set(tweets, "SB", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(SB)"].append(data["Naive Bayes"][0])
f1_data["NB(SB)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(SB)"].append(data["SVM"][0])
f1_data["SVM(SB)"].append(data["SVM"][3])
accuracy_data["MaxEnt(SB)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(SB)"].append(data["Maximum Entropy"][3])
data = train_subjectivity_and_test_on_feature_set(tweets, "SC", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(SC)"].append(data["Naive Bayes"][0])
f1_data["NB(SC)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(SC)"].append(data["SVM"][0])
f1_data["SVM(SC)"].append(data["SVM"][3])
accuracy_data["MaxEnt(SC)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(SC)"].append(data["Maximum Entropy"][3])
data = train_polarity_and_test_on_feature_set(tweets, "PA", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(PA)"].append(data["Naive Bayes"][0])
f1_data["NB(PA)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(PA)"].append(data["SVM"][0])
f1_data["SVM(PA)"].append(data["SVM"][3])
accuracy_data["MaxEnt(PA)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(PA)"].append(data["Maximum Entropy"][3])
data = train_polarity_and_test_on_feature_set(tweets, "PB", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(PB)"].append(data["Naive Bayes"][0])
f1_data["NB(PB)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(PB)"].append(data["SVM"][0])
f1_data["SVM(PB)"].append(data["SVM"][3])
accuracy_data["MaxEnt(PB)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(PB)"].append(data["Maximum Entropy"][3])
data = train_polarity_and_test_on_feature_set(tweets, "PC", sentimentvalues, reduce_dataset=i * 0.01)
print "DATA -- ", data
accuracy_data["NB(PC)"].append(data["Naive Bayes"][0])
f1_data["NB(PC)"].append(data["Naive Bayes"][3])
accuracy_data["SVM(PC)"].append(data["SVM"][0])
f1_data["SVM(PC)"].append(data["SVM"][3])
accuracy_data["MaxEnt(PC)"].append(data["Maximum Entropy"][0])
f1_data["MaxEnt(PC)"].append(data["Maximum Entropy"][3])
out = open("incremental_acc" + str(i), "wb")
pickle.dump(accuracy_data, out)
out = open("incremental_f1" + str(i), "wb")
pickle.dump(f1_data, out)
plotting.plot_temporal_sentiment(accuracy_data, filename="incremental_accuracy")
plotting.plot_temporal_sentiment(f1_data, filename="incremental_f1")
pickle.dump(data, open("topically_aggregated_polarity", "wb"))
def preprocess_temporal_dataset():
tweetlines = utils.get_dataset(utils.complete_datasets[3])
tweets = []
for line in tweetlines:
if len(line) > 1:
tweets.append(tweet.to_tweet(line))
tweets = preprocessing.preprocess_tweets(tweets)
sentiments = lexicon.perform_google_sentiment_lexicon_lookup(tweets)
pickle.dump(sentiments, open("temporal_sentiments", "wb"))
pickle.dump(tweets, open("temporal_tweets2", "wb"))
if __name__ == "__main__":
datasetnr = 3
tweets = utils.get_pickles(datasetnr)
sentimentvalues = feat_utils.get_sentiment_values(datasetnr)
tweets = preprocessing.remove_link_classes(tweets)
tweets = preprocessing.lower_case(tweets)
tweets = preprocessing.remove_specialchars_round2(tweets)
train_subjectivity_and_test_on_feature_set(tweets, "SA", datasetnr)
train_subjectivity_and_test_on_feature_set(tweets, "SB", datasetnr)
train_subjectivity_and_test_on_feature_set(tweets, "SC", sentimentvalues)
train_polarity_and_test_on_feature_set(tweets, "PA", datasetnr)
train_polarity_and_test_on_feature_set(tweets, "PB", datasetnr)
train_polarity_and_test_on_feature_set(tweets, "PC", sentimentvalues)
