def getEmoticons(tokenised_tweet, emo_dict):
feature_emoticons = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word in tokenised_tweet:
if isEmoticon(word, emo_dict):
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], word, 1)
return feature_emoticons
def getUniqueFeaturesForClass(processed_tweets_list,
categorywide_unique_features,
features_used=FEATURES_SA_DEFAULT):
'''
Params:
processed_tweets_list: [] of tweets of class
categorywide_unique_features: {} to keep track of unique features of a category e.g. apple, google,...
features_used: [] of features in use from CS4242_Assg2.constants
Return:
class_unique_features: { FEATURE_TYPE_... : {feature: value}}
'''
class_unique_features = {}
for feature in features_used:
# TODO: whitelist
if feature == FEATURE_SA_REPLIES or feature == FEATURE_SA_TEMPORAL:
continue
class_unique_features[feature] = {}
for processed_tweet in processed_tweets_list:
for feature in features_used:
# TODO: whitelist
if feature == FEATURE_SA_REPLIES or feature == FEATURE_SA_TEMPORAL:
continue
val_dict = processed_tweet[TWEET_FEATURES][feature][FEATURE_VALUE]
for key, val in val_dict.iteritems():
addToCountDict(categorywide_unique_features[feature], key, 1)
addToCountDict(class_unique_features[feature], key, 1)
return class_unique_features
def getCapitalisedText(tokenised_tweet):
feature_caps = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word in tokenised_tweet:
# split_words = removePunctuationsAndNumbers(word)
# for w in split_words:
if isFullCaps(word):
feature_caps[FEATURE_COUNT] += 1
addToCountDict(feature_caps[FEATURE_VALUE], word, 1)
return feature_caps
def getHashTags_SA(json_data, tokenised_tweet_no_filter):
feature_ht = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for ht in json_data['entities']['hashtags']:
feature_ht[FEATURE_COUNT] += 1
addToCountDict(feature_ht[FEATURE_VALUE], "#HT_" + ht['text'], 1)
for ht in getUncapturedHashtags(tokenised_tweet_no_filter):
feature_ht[FEATURE_COUNT] += 1
addToCountDict(feature_ht[FEATURE_VALUE], "#HT_" + ht, 1)
return feature_ht
def getPosFeatures(pos_tagged_tweet):
'''
{
FEATURE_COUNT: number of POS (a,n,v,r,other) features in tweet
FEATURE_VALUE: {pos_tag: number of pos_tag features in tweet}
}
'''
feature_pos = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word, pos in pos_tagged_tweet.iteritems():
if pos != POS_OTHERS:
feature_pos[FEATURE_COUNT] += 1
addToCountDict(feature_pos[FEATURE_VALUE], "POS_TAG_%s" % pos, 1)
return feature_pos
def extractTextFeatures(json_data):
tweet_words = {}
# Remove unnecessary tokens
tweet_text = filterIrrelevantTokens(json_data)
unicode_normalized_tweet = unicodedata.normalize(
'NFKD', tweet_text).encode('ascii', 'ignore').lower()
tweet_wordlist = normalizer.normalizeTweet(unicode_normalized_tweet)
# tweet wordlist = ['word', 'word2'] etc
for word in tweet_wordlist:
stripped_punct_num_word = removePunctuationsAndNumbers(word)
for w in stripped_punct_num_word:
if w not in stopwords.words('english'):
w2 = stemmer.stem(w)
addToCountDict(tweet_words, w2, 1)
return tweet_words
def getUserMentions_SA(json_data, tokenised_tweet_no_filter):
'''
Get user ids of users mentioned in tweet, include tweeter of tweet
'''
feature_obj = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for um in json_data['entities']['user_mentions']:
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE], "@UM_" + um['id_str'], 1)
# addToCountDict(feature_obj[FEATURE_VALUE], "@UM_" + str(json_data['user']['id']), 1)
for um in getUncapturedUserMentions(tokenised_tweet_no_filter):
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE], "@UM_" + um, 1)
for um in feature_obj[FEATURE_VALUE]:
feature_obj[FEATURE_VALUE][um] = 1
return feature_obj
def getText(tokenised_tweet, negation_flags, use_negation=False):
feature_text = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word in tokenised_tweet:
negated = False
if word in negation_flags and negation_flags[word]:
negated = True
if isEmoticon(word, emo_dict):
continue
is_slang = isSlang(word, sd)
if is_slang:
translated_slang = translateSlangWord(word, sd)
tokenised_slang = tokenizer.tokenize(translated_slang)
if len(tokenised_slang) == 1: # translated slang
split_words = tokenised_slang
is_slang = False
else: # don't remove punctuation/numbers
split_words = [word]
# print split_words
if not is_slang:
split_words = removePunctuationsAndNumbers(word)
for w in split_words:
w = w.lower()
if w in stopwords.words('english'): # skip stopwords
continue
if not is_slang: # stem if its not slang
w = stemmer.stem(w)
if isFullCaps(word): # preserve case for words in full
w = w.upper()
if use_negation and negated:
w = "NOT_%s" % w
# print w
feature_text[FEATURE_COUNT] += 1
addToCountDict(feature_text[FEATURE_VALUE], w, 1)
# TODO: should we use presence? results will drop abit
# for text in feature_text[FEATURE_VALUE]:
# feature_text[FEATURE_VALUE][text] = 1
return feature_text
def getPolarityPosCount(pos_tagged_tweet, tweet_word_polarity):
'''
{
FEATURE_COUNT: number of text features of POS (a,n,v,r) with polarity in tweet
FEATURE_VALUE: {polarity type: number of text features of POS (a,n,v,r) with polarity in tweet}
}
'''
feature_pol = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word, pos in pos_tagged_tweet.iteritems():
# word_polarity = getPolarityFromSWN(word, pos, swn) # empty string if word not found
if pos == POS_OTHERS:
continue
word_polarity = getPolarityOfWord(word, tweet_word_polarity)
if word_polarity == POLARITY_POSITIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_PoS_POS", 1)
if word_polarity == POLARITY_NEGATIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_PoS_NEG", 1)
if word_polarity == POLARITY_NEUTRAL:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_PoS_NEU", 1)
return feature_pol
def getEmoticonPolarityCount(tokenised_tweet, emo_dict):
feature_emoticons = {
FEATURE_COUNT: 0,
FEATURE_VALUE: {
'EMO_V_POS': 0,
'EMO_POS': 0,
'EMO_NEU': 0,
'EMO_NEG': 0,
'EMO_V_NEG': 0
}
}
for word in tokenised_tweet:
if word in emo_dict:
polarity = getEmoticonPolarity(word, emo_dict)
if polarity == POLARITY_VERY_POSITIVE:
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], 'EMO_V_POS',
1)
elif polarity == POLARITY_POSITIVE:
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], 'EMO_POS', 1)
elif polarity == POLARITY_NEUTRAL:
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], 'EMO_NEU', 1)
elif polarity == POLARITY_NEGATIVE:
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], 'EMO_NEG', 1)
elif polarity == POLARITY_VERY_NEGATIVE:
feature_emoticons[FEATURE_COUNT] += 1
addToCountDict(feature_emoticons[FEATURE_VALUE], 'EMO_V_NEG',
1)
# addToCountDict(feature_emoticons[FEATURE_VALUE], word, 1)
# print feature_emoticons
return feature_emoticons
def getCapitalisedTextPolarityCount(cap_text, tweet_word_polarity):
feature_pol = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word in cap_text[FEATURE_VALUE]:
word_polarity = getPolarityOfWord(word, tweet_word_polarity)
if word_polarity == POLARITY_POSITIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_CAPS_POS", 1)
if word_polarity == POLARITY_NEGATIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_CAPS_NEG", 1)
if word_polarity == POLARITY_NEUTRAL:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARTITY_CAPS_NEU", 1)
return feature_pol
def getTwitterTokenCount(json_data, tokenised_tweet_no_filter):
feature_obj = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for twitter_token, entity in json_data['entities'].iteritems():
for value in entity:
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE],
"TWITTER_TOKEN_" + twitter_token, 1)
for token in tokenised_tweet_no_filter:
if re.match(r"^#", token):
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE],
'TWITTER_TOKEN_hashtags', 1)
if re.match(r"^@", token):
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE],
'TWITTER_TOKEN_user_mentions', 1)
if isValidUrl(token):
feature_obj[FEATURE_COUNT] += 1
addToCountDict(feature_obj[FEATURE_VALUE], 'TWITTER_TOKEN_urls', 1)
# print feature_obj
return feature_obj
def getPolarityTextCount(pos_tagged_tweet, tweet_word_polarity):
'''
{
FEATURE_COUNT: number of text features with polarity
FEATURE_VALUE: {text: polarity score of text}
}
'''
feature_pol = {FEATURE_COUNT: 0, FEATURE_VALUE: {}}
for word, pos in pos_tagged_tweet.iteritems():
word_polarity = getPolarityOfWord(word, tweet_word_polarity)
if word_polarity == POLARITY_POSITIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARITY_TEXT_POS", 1)
elif word_polarity == POLARITY_NEGATIVE:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARITY_TEXT_NEG", 1)
elif word_polarity == POLARITY_NEUTRAL:
feature_pol[FEATURE_COUNT] += 1
addToCountDict(feature_pol[FEATURE_VALUE], "POLARITY_TEXT_NEU", 1)
return feature_pol
def calculateChi2ValuesSA(unique_feat, feature_set, sample_size_set):
'''
Calculate Chi2 values of all unique features using lecture's formula
Params:
unique_feat: [feat] of unique features
feature_set: {senti_type: features} of features from each sentiment
sample_size_set: {senti_type:sample_tweet_count} of each sentiment
Returns:
chi2_val_list: {senti_type: {feat: chi2_val}}of calculated chi2 values
'''
# print 'using calculateChi2Values...'
sentiment_feature_set = {}
for target_sentiment in feature_set.iterkeys():
target_senti_feat_dict = feature_set[target_sentiment]
target_senti_sample_size = sample_size_set[target_sentiment]
# merge the other sentiment feature dictionaries
other_senti_feat_dict = {}
other_senti_sample_size = 0
for sentiment in feature_set.iterkeys():
if sentiment != target_sentiment:
for item, value in feature_set[sentiment].iteritems():
addToCountDict(other_senti_feat_dict, item, value)
other_senti_sample_size += sample_size_set[sentiment]
target_sent_feat_set = {}
target_sent_feat_set['target_feat_dict'] = target_senti_feat_dict
target_sent_feat_set['target_sample_size'] = sum(
target_senti_feat_dict.itervalues())
target_sent_feat_set['other_feat_dict'] = other_senti_feat_dict
target_sent_feat_set['other_sample_size'] = sum(
other_senti_feat_dict.itervalues())
sentiment_feature_set[target_sentiment] = target_sent_feat_set
# target_senti_sample_size = sum(target_senti_feat_dict.itervalues())
# other_senti_sample_size = sum(other_senti_feat_dict.itervalues())
# print target_sent_feat_set['target_sample_size'], target_sent_feat_set['other_sample_size']
# print sum(unique_feat.itervalues())
chi2_val_dict = {}
for target_sentiment in feature_set.iterkeys():
target_sent_feat_set = sentiment_feature_set[target_sentiment]
target_senti_feat_dict = target_sent_feat_set['target_feat_dict']
target_senti_sample_size = target_sent_feat_set['target_sample_size']
other_senti_feat_dict = target_sent_feat_set['other_feat_dict']
other_senti_sample_size = target_sent_feat_set['other_sample_size']
chi2_word_val = {}
for f in target_senti_feat_dict.iterkeys():
# calculate chi2
A = 0 # num of tweets in target cat that contains feature
B = 0 # num of tweets not in target cat that contains feature
C = 0 # num of tweets in target cat that doesnt contains feature
D = 0 # num of tweets not in target cat that doesnt contains feature
if f in target_senti_feat_dict:
A = target_senti_feat_dict[f]
if f in other_senti_feat_dict:
B = other_senti_feat_dict[f]
C = target_senti_sample_size - A
D = other_senti_sample_size - B
chi2_word_val[f] = calculateChi2(A, B, C, D)
# print f,A,B,C,D, chi2_word_val[f]
# break
chi2_val_dict[target_sentiment] = chi2_word_val
return chi2_val_dict
def getKeyInfoForClassifier(filename,
categories_list,
groundtruth_list,
features=FEATURES_DEFAULT):
'''
Extracts all features from input files.
Params:
filename: input DATA file
categories_list: list of categories, e.g ['Apple', 'Google', 'Twitter']
groundtruth_list: list of groundtruths, e.g [{CATEGORY: category, POLARITY: polarity, TWEET_ID: tweetid}]
features (optional): [FEATURE_TEXT, FEATURE_HASHTAG, FEATURE_GEOINFO, FEATURE_FOLLOWED_CATEGORIES,
FEATURE_USER, FEATURE_USER_MENTIONS]
Returns:
{
'category' : {
POSITIVE:{
PROCESSED_TWEETS : [{
TWEET_FULL: This was a triumph,
TWEET_FEATURES: {
FEATURE_TEXT: {} ,
FEATURE_GEOLOCATION : str
}
}],
FEATURES: {
FEATURE_TEXT: {} ...
}
},
NEGATIVE:{
PROCESSED_TWEETS : [{
TWEET_FULL: This was a triumph,
TWEET_FEATURES: {
FEATURE_TEXT: {} ,
FEATURE_GEOLOCATION : str
}
}],
FEATURES: {
FEATURE_TEXT: {} ...
FEATURE_HASHTAG: []
}
},
UNIQUE_FEATURES: {
FEATURE_TEXT: {} , ...
}
}
}
'''
returnmap = {}
for category in categories_list:
# category wide variables
processed_tweets_list = []
unique_features_map = {}
positive_processed_tweet_list = []
negative_processed_tweet_list = []
positive_features_map = {}
negative_features_map = {}
# Initialize unique feature maps within category
for feature in features:
if feature == FEATURE_TEXT:
feature_text_unique = {}
pos_feature_text_unique = {}
neg_feature_text_unique = {}
elif feature == FEATURE_HASHTAG:
feature_hashtag_unique = {}
pos_feature_hashtag_unique = {}
neg_feature_hashtag_unique = {}
elif feature == FEATURE_GEOINFO:
feature_geoinfo_unique = {}
pos_feature_geoinfo_unique = {}
neg_feature_geoinfo_unique = {}
elif feature == FEATURE_FOLLOWED_CATEGORIES:
feature_followed_cat_unique = {}
pos_feature_followed_cat_unique = {}
neg_feature_followed_cat_unique = {}
elif feature == FEATURE_USER:
feature_user_unique = {}
pos_feature_user_unique = {}
neg_feature_user_unique = {}
elif feature == FEATURE_USER_MENTIONS:
feature_usermentions_unique = {}
pos_feature_usermentions_unique = {}
neg_feature_usermentions_unique = {}
elif feature == FEATURE_CATEGORY:
feature_category_unique = {}
pos_feature_category_unique = {}
neg_feature_category_unique = {}
# Extract & Process Features
with codecs.open(filename, encoding='cp1252') as k:
for index, line in enumerate(k):
json_data = json.loads(line, encoding='cp1252')
tweet_keyinfo = extractFeaturesFromTweet(
json_data, categories_list, features, category)
if groundtruth_list[index][CATEGORY] == category:
positive_processed_tweet_list.append(tweet_keyinfo)
else:
negative_processed_tweet_list.append(tweet_keyinfo)
processed_tweets_list.append(tweet_keyinfo)
# check unique tweet_keyinfo across positive, negative and all
for tweet_keyinfo in positive_processed_tweet_list:
for feature in features:
if feature == FEATURE_TEXT:
text_count_dict = tweet_keyinfo[TWEET_FEATURES][
FEATURE_TEXT]
for key, count in text_count_dict.iteritems():
addToCountDict(pos_feature_text_unique, key, 1)
elif feature == FEATURE_HASHTAG:
for hashtag in tweet_keyinfo[TWEET_FEATURES][
FEATURE_HASHTAG]:
addToCountDict(pos_feature_hashtag_unique, hashtag, 1)
elif feature == FEATURE_GEOINFO:
if tweet_keyinfo[TWEET_FEATURES][FEATURE_GEOINFO] != '':
addToCountDict(
pos_feature_geoinfo_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_GEOINFO], 1)
elif feature == FEATURE_FOLLOWED_CATEGORIES:
# TODO: Consider implementing
pass
elif feature == FEATURE_USER:
addToCountDict(pos_feature_user_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_USER],
1)
elif feature == FEATURE_USER_MENTIONS:
for usermention in tweet_keyinfo[TWEET_FEATURES][
FEATURE_USER_MENTIONS]:
addToCountDict(pos_feature_usermentions_unique,
usermention, 1)
elif feature == FEATURE_CATEGORY:
for item in tweet_keyinfo[TWEET_FEATURES][
FEATURE_CATEGORY]:
addToCountDict(pos_feature_category_unique, item, 1)
# debug files will be written only if settings.DEBUG_CODE = True
# writeDebugCountDictToFile("%s_pos_feature_text_unique.txt" % category, pos_feature_text_unique)
# writeDebugCountDictToFile("%s_pos_feature_hashtag_unique.txt" % category, pos_feature_hashtag_unique)
# writeDebugCountDictToFile("%s_pos_feature_geoinfo_unique.txt" % category, pos_feature_geoinfo_unique)
# writeDebugCountDictToFile("%s_pos_feature_user_unique.txt" % category, pos_feature_user_unique)
# writeDebugCountDictToFile("%s_pos_feature_usermentions_unique.txt" % category, pos_feature_usermentions_unique)
for tweet_keyinfo in negative_processed_tweet_list:
for feature in features:
if feature == FEATURE_TEXT:
text_count_dict = tweet_keyinfo[TWEET_FEATURES][
FEATURE_TEXT]
for key, count in text_count_dict.iteritems():
addToCountDict(neg_feature_text_unique, key, 1)
elif feature == FEATURE_HASHTAG:
for hashtag in tweet_keyinfo[TWEET_FEATURES][
FEATURE_HASHTAG]:
addToCountDict(neg_feature_hashtag_unique, hashtag, 1)
elif feature == FEATURE_GEOINFO:
addToCountDict(
neg_feature_geoinfo_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_GEOINFO], 1)
elif feature == FEATURE_FOLLOWED_CATEGORIES:
# TODO: Consider implementing
pass
elif feature == FEATURE_USER:
addToCountDict(neg_feature_user_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_USER],
1)
elif feature == FEATURE_USER_MENTIONS:
for usermention in tweet_keyinfo[TWEET_FEATURES][
FEATURE_USER_MENTIONS]:
addToCountDict(neg_feature_usermentions_unique,
usermention, 1)
elif feature == FEATURE_CATEGORY:
for item in tweet_keyinfo[TWEET_FEATURES][
FEATURE_CATEGORY]:
addToCountDict(neg_feature_category_unique, item, 1)
for tweet_keyinfo in processed_tweets_list:
for feature in features:
if feature == FEATURE_TEXT:
text_count_dict = tweet_keyinfo[TWEET_FEATURES][
FEATURE_TEXT]
for key, count in text_count_dict.iteritems():
addToCountDict(feature_text_unique, key, 1)
elif feature == FEATURE_HASHTAG:
for hashtag in tweet_keyinfo[TWEET_FEATURES][
FEATURE_HASHTAG]:
addToCountDict(feature_hashtag_unique, hashtag, 1)
elif feature == FEATURE_GEOINFO:
addToCountDict(
feature_geoinfo_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_GEOINFO], 1)
elif feature == FEATURE_FOLLOWED_CATEGORIES:
# TODO: Consider implementing
pass
elif feature == FEATURE_USER:
addToCountDict(feature_user_unique,
tweet_keyinfo[TWEET_FEATURES][FEATURE_USER],
1)
elif feature == FEATURE_USER_MENTIONS:
for usermention in tweet_keyinfo[TWEET_FEATURES][
FEATURE_USER_MENTIONS]:
addToCountDict(feature_usermentions_unique,
usermention, 1)
elif feature == FEATURE_CATEGORY:
for item in tweet_keyinfo[TWEET_FEATURES][
FEATURE_CATEGORY]:
addToCountDict(feature_category_unique, item, 1)
for feature in features:
if feature == FEATURE_TEXT:
unique_features_map[FEATURE_TEXT] = feature_text_unique
positive_features_map[FEATURE_TEXT] = pos_feature_text_unique
negative_features_map[FEATURE_TEXT] = neg_feature_text_unique
elif feature == FEATURE_HASHTAG:
unique_features_map[FEATURE_HASHTAG] = feature_hashtag_unique
positive_features_map[
FEATURE_HASHTAG] = pos_feature_hashtag_unique
negative_features_map[
FEATURE_HASHTAG] = neg_feature_hashtag_unique
elif feature == FEATURE_GEOINFO:
unique_features_map[FEATURE_GEOINFO] = feature_geoinfo_unique
positive_features_map[
FEATURE_GEOINFO] = pos_feature_geoinfo_unique
negative_features_map[
FEATURE_GEOINFO] = neg_feature_geoinfo_unique
elif feature == FEATURE_FOLLOWED_CATEGORIES:
unique_features_map[
FEATURE_FOLLOWED_CATEGORIES] = feature_followed_cat_unique
positive_features_map[
FEATURE_FOLLOWED_CATEGORIES] = pos_feature_followed_cat_unique
negative_features_map[
FEATURE_FOLLOWED_CATEGORIES] = neg_feature_followed_cat_unique
elif feature == FEATURE_USER:
unique_features_map[FEATURE_USER] = feature_user_unique
positive_features_map[FEATURE_USER] = pos_feature_user_unique
negative_features_map[FEATURE_USER] = neg_feature_user_unique
elif feature == FEATURE_USER_MENTIONS:
unique_features_map[
FEATURE_USER_MENTIONS] = feature_usermentions_unique
positive_features_map[
FEATURE_USER_MENTIONS] = pos_feature_usermentions_unique
negative_features_map[
FEATURE_USER_MENTIONS] = neg_feature_usermentions_unique
elif feature == FEATURE_CATEGORY:
unique_features_map[FEATURE_CATEGORY] = feature_category_unique
positive_features_map[
FEATURE_CATEGORY] = pos_feature_category_unique
negative_features_map[
FEATURE_CATEGORY] = neg_feature_category_unique
returnmap[category] = {}
returnmap[category][POSITIVE] = {
PROCESSED_TWEETS: positive_processed_tweet_list,
FEATURES: positive_features_map
}
returnmap[category][NEGATIVE] = {
PROCESSED_TWEETS: negative_processed_tweet_list,
FEATURES: negative_features_map
}
returnmap[category][UNIQUE_FEATURES] = unique_features_map
return returnmap
