def predict(n, url):
article = Article(url, language="en")
article.download()
article.parse()
article.nlp()
content = pre_process_article(str(article.text))
title = pre_process_article(str(article.title))
tf_idf(content, title, n)
def runEngine(tokenizer_type, address):
print("RUNNING ENGINE ...")
inverted_index = load_obj("INVERTED INDEX " + tokenizer_type + "_" +
address)
tfidfs = load_obj("CHAMPIONS " + tokenize_type + "_" + address)
query = ""
while True:
query = input("QUERY :>")
if query == "!q": break
queryToken = query.split(" ")
queryw = tf_idf(queryToken, inverted_index, False)
h = Heap()
for i in range(len(tfidfs)):
if bool(set(queryToken) & set(tfidfs[i].keys())):
sim = querySimilarity(queryw, tfidfs[i])
if sim != 0:
h.addnSort([i, sim])
k = 10
result = h.getFirstK(k)
titles = fetch_column(address, 'title')
for i in range(k):
print(titles[result[i][0]][::-1])
def getChampions(tokens, inverted_index):
tfidfs = []
for i in range(len(tokens)):
tfidfs.append(tf_idf(tokens[i], inverted_index, False))
champions_term = {}
champions_list = {}
# TODO: Optimize this mess
for term in inverted_index:
champions_term[term] = [None] * inverted_index[term][0]
for i in range(0, inverted_index[term][0]):
champions_term[term][i] = tfidfs[inverted_index[term][i + 1]][term]
for term in champions_term:
y = list(
zip(*heapq.nlargest(10,
enumerate(champions_term[term]),
key=operator.itemgetter(1))))[0]
champions_list[term] = list(y)
for term in champions_list:
l = min(10, len(champions_list[term]))
for i in range(l):
champions_list[term][i] = inverted_index[term][
champions_list[term][i] + 1]
return champions_list
def make_tf_matrix(self, doc_list):
tf_matrix = []
for doc in doc_list:
row = [0] * self.count
for word in doc.split(' '):
row[self.index[word]] = tfidf.tf_idf(word, doc, doc_list)
tf_matrix.append(row)
return tf_matrix
def generate_classifier_model(tokenized_sentences, features):
model = []
import tfidf
tfidf.cache_enabled(True)
for sentence in tokenized_sentences:
tweet_model = {}
for i, feature in enumerate(features):
tfidfv = tfidf.tf_idf(feature, sentence, tokenized_sentences)
if tfidfv > 0:
tweet_model[i] = tfidfv
model.append(tweet_model)
return model
def compute_document_score_by_term(term, tags=None):
# documents = find_documents(term)
scores = tfidf.tf_idf(term, tags)
# start_time = time.time()
# res = tfidf.terms_collection.aggregate([
# {'$match': {'term': term}},
# {
# '$group': {
# '_id': "$doc",
# 'count': {'$sum': 1}
# }
# }
# ])
# for doc in res:
# scores[doc['_id']] = tfidf.tf_idf(term, doc['_id'], doc['count'])
# print "Find Score Time = ", str((time.time() - start_time))
# for doc in documents:
# scores[doc] = tfidf.tf_idf(term, doc)
return scores
def callback(self):
print("GOT TOKEN ", self.text.get("1.0", END))
queryToken = self.text.get("1.0", END).replace("\n", "").split(" ")
queryToken = normalize_query(queryToken)
print(queryToken)
queryw = tf_idf(queryToken, self.inverted_index, False)
h = Heap()
for i in range(len(self.tfidfs)):
if bool(set(queryToken) & set(self.tfidfs[i].keys())):
sim = querySimilarity(queryw, self.tfidfs[i])
if sim != 0:
h.addnSort([i, sim])
k = 10
result = h.getFirstK(k)
k = min(len(result), k)
print(k)
for i in range(k):
print(self.titles[result[i][0]])
self.mylist.delete(i)
self.mylist.insert(i, self.titles[result[i][0]])
def get_bag_of_words_labels(preprocessed_records, args):
"""Gets the labels for the bag of words. A label can be a a single important word, a collocation of two important
words, or a set of synonyms of a word.
Params:
- preprocessed_records (pyspark.rdd.RDD): The tokenized, lemmatized, lowercase records
- ars (argparse.Namespace): The command-line arguments passed to the program
Returns:
- bag_of_words_labels (list<str|tuple<str>>): The labels of the bag of words created
"""
reformatted_records = preprocessed_records.map(lambda record: (record['id'], record['preprocessed_record']))
frequent_collocations = wordcount.extract_collocations(reformatted_records, args.num_collocations,
args.collocation_window)
tf_idf_scores = tfidf.tf_idf(reformatted_records)
# Pyspark technically ends here - the rest is processed on master node
important_words = tfidf.extract_important_words(tf_idf_scores, args.num_words, False)
# important_words_with_counts = synsets.add_word_counts(important_words, frequent_words)
synset_dict = synsets.generate_syn_set(important_words)
words_and_collocations = wordcount.merge_collocations_with_wordlist(frequent_collocations, important_words)
# Merge words, collocations and synsets
bag_of_words_labels = list()
for item in words_and_collocations:
if " " in item: # item is a collocation
bag_of_words_labels.append(item)
elif item in synset_dict: # item is an important word
synset = synset_dict[item]
if len(synset) == 1: # synset only contains the word itself
bag_of_words_labels.append(item)
else: # synset contains multiple words
synset = [word.encode('utf-8') for word in synset[1:]]
bag_of_words_labels.append(synset)
# Save bag of words labels to single text file
with open("bag_of_words_labels.json", "w") as bow_file:
json.dump(bag_of_words_labels, bow_file)
return bag_of_words_labels
def main_(set_,v=None):
data_pool = os.listdir(PATH+YEAR) #準備全樣本
s = sorted(list(set_)) #部分樣本代碼排序
res = []
if v is None:
lst = []
#取開頭4個字元 轉成int 以便搜尋
for i in range(len(data_pool)):
data_pool[i] = int(data_pool[i][0:4])
for key in s:
res = binary_search(data_pool, key) #二分法搜尋
lst.append(res) #return index 【搜尋特定產業CSV】
for w in lst:
write_txt(data_pool[w], YEAR)
# #一次取2個進行比較
for idx in lst:
for idx_ in lst:
print(str(data_pool[idx])+"<--->"+str(data_pool[idx_]))
s1 = get_txt(data_pool[idx])
s2 = get_txt(data_pool[idx_])
j1 = jieba_(s1)
j2 = jieba_(s2)
vector = tf_idf(j1,j2)
cos = get_cos(vector[0], vector[1])
with open("化學"+YEAR+".txt", "a+" )as f:
f.write(str(data_pool[idx])+"<--->"+str(data_pool[idx_])+" : "+ str(cos)+"\n")
import re
import tfidf
#import re
#pattern = '[0-9]+[\t]+(.+)[\t]+[0-9]+'
#p = re.compile(pattern)
#m = p.search("\"1 theo walcott is still shit, watch rafa and johnny deal with him on saturday. 1\"")
#content = m.group(1)
#print(content)
# test tf idf
#i hava hava a dog
#i like
# i :1/2 ,have:2/2, a :1/2 dog:1/2
# i:1 like:1
#tf-idf i:0 have:1 a:1/2 dog:1/2
#i:0 like:1
res = tfidf.tf_idf([['i', 'have', 'have', 'a', 'dog'], ['i', 'like']])
tfidf.dump_tfidf_json(res, 'res.json')
raw_data, raw_labels = load_data_from_file()
print(
f'[main] Raw data dims - tweets: {len(raw_data)} / labels: {len(raw_labels)}'
)
# Pre-process tweets & labels:
corpus, labels = preprocess(raw_data, raw_labels, stemmer_flag)
store_processed_data(corpus, labels) # save processed data to file
print(
f'[main] Processed data dims - corpus: {len(corpus)} / labels: {len(labels)}'
)
# Compute TF-IDF or Graph of Words
if method == 'TFIDF':
X = tf_idf(corpus)
elif method == 'GOW':
X = gow(corpus)
else:
raise ValueError(f'[main] Invalid method: {method}')
# Split the data
X_train, X_test, y_train, y_test = train_test_split(
X,
labels,
test_size=test_size,
stratify=labels,
random_state=RANDOM_STATE,
)
print(
f'[main] Training/test set dimensions: {len(y_train)}/{len(y_test)}\n')
algo = 'MLP'
method = 'TFIDF'
test_size = .2
estimator = get_classifier(algo)
X, y = load_processed_data()
# create multi-label like settings
handles = get_handles()
y = label_binarize(y, classes=[handles[0], handles[1], handles[2]])
n_classes = y.shape[1]
# Compute TF-IDF or Graph of Words
if method == 'TFIDF':
X = tf_idf(X)
elif method == 'GOW':
X = gow(X)
else:
raise ValueError(f'[main] Invalid method: {method}')
# Split into training and test
X_train, X_test, Y_train, Y_test = train_test_split(
X, y, test_size=test_size, random_state=RANDOM_STATE)
X_train, X_test = normalize_X(X_train, X_test)
# Load the best model from GridSearchCV()
model = load_model(algo, method, 'tuning_random')
params = model.best_params_
print(f'Best model parameters: \n{params}\n')
#params.update({'n_jobs': n_jobs},)
import tool
import tfidf
filename = 'tweet.txt'
content_list = tool.read_tweet_content(filename)
print(content_list[0])
print('- - ' * 50)
print(content_list[1])
stem_list = tfidf.preprocess(content_list)
print('- - ' * 50)
print(stem_list[0])
print('- - ' * 50)
print(stem_list[1])
token_2d_list = tfidf.tokenization(stem_list)
print(token_2d_list[0])
print(token_2d_list[1])
print('res')
res = tfidf.tf_idf(token_2d_list)
print('dump')
tfidf.dump_tfidf_json(res, 'tfidf.json')
from tfidf import tf_idf
from similarity import cosine_similarity
import os
copyrighted_works='/Users/sm.marouzigmail.com/Documents/job/wattpad/library/'
user_text_file = '/Users/sm.marouzigmail.com/Documents/job/wattpad/test/test.txt'
# read the content of new uploaded text
try:
with open(user_text_file, "r", encoding='iso-8859-1') as ifile:
raw_text = ifile.read() # raw text of new uploaded file
except:
with open(user_text_file, "r", encoding='utf-8') as ifile:
raw_text = ifile.read() # raw text of new uploaded file
tfs_dict1 = tf_idf(raw_text)
library_path=os.listdir(copyrighted_works)
# list of text files
files = [file for file in library_path if file.endswith('.txt')]
for file in files:
# read text data for each copyrighted content
# extract TF_IDF terms and values
# calculate the similarity between this vector and uploaded text file vector
full_path_file = os.path.join(copyrighted_works, file)
# read raw texts one by one from copyright library path
try:
parser = argparse.ArgumentParser(description='Generate Inverted Index')
parser.add_argument('address', help='csv address', action='store')
parser.add_argument('tokenize',
help='type of tokenization: simple or pro',
default='simple')
args = parser.parse_args()
csv_address = args.address
tokenize_type = args.tokenize
print("READING ", csv_address)
contents = fetch_column(csv_address, 'content')
tokens = tokenize(tokenize_type, contents)
print("TOKENIZING ", tokenize_type)
if tokenize_type == 'pro':
tokens = normilize(tokens)
tokens = stem_list(tokens)
print("BUILDING DICTIONARY AND INVERTED INDEX")
dictionary = build_dictionary(tokens)
inverted_index = build_inverted_index(tokens, dictionary)
print("WRITING INDEX AND SCORES")
writeObj("INVERTED INDEX", inverted_index, tokenize_type, csv_address)
tfidfs = []
for i in range(len(tokens)):
tfidfs.append(tf_idf(tokens[i], inverted_index, True))
writeObj("TFIDF", tfidfs, tokenize_type, csv_address)
def indices(n):
for i, j in product(range(n), range(n)):
if j < i: yield i, j
if __name__ == '__main__':
limit = 1000
tfidf_thresh = 0.02
vocab = open('data/vocab.txt', 'r').read().split('\n')
embeddings = np.load('data/embeddings.npy', allow_pickle=False)
tok2id = {tok: i for i, tok in enumerate(vocab)}
tf, idf = tf_idf(stream_tokens(limit))
# Compute TF-IDFs
# and create doc representations
print('Creating doc matrices...')
pmid_idx = {}
mats = []
for pmid, toks in tqdm(stream_tokens(limit)):
ems = []
for tok in set(toks):
tfidf = tf[pmid][tok] * idf[tok]
if tfidf >= tfidf_thresh:
tid = tok2id.get(tok)
if tid is None:
continue
em = embeddings[tid]
def make_data():
f = open('cl.sents.txt', 'r', encoding='utf-8')
texts = f.readlines()
f.close()
docs = {}
i = 0
for el in texts:
docs[i] = el
i = i + 1
return docs
docs = make_data()
index = tf_idf(docs)
def make_data2():
f = open('sents.txt', 'r', encoding='utf-8')
texts = f.readlines()
f.close()
docs2 = {}
i = 0
for el in texts:
docs2[i] = el
i = i + 1
return docs2
docs2 = make_data2()
:return: A dictionary mapping words in wordlist to their word counts
"""
wordlist_with_counts = dict()
for word in wordlist:
if word in word_counts:
wordlist_with_counts[word] = word_counts[word]
else:
print(word, " not in word_counts dictionary but in given wordlist")
return wordlist_with_counts
# End of add_word_counts()
if __name__ == "__main__":
args = wordcount.parse_arguments()
records = wordcount.load_records(args.file, False)
records = wordcount.preprocess_records(records)
frequent_words = wordcount.extract_frequent_words(records,
args.num_words * 10,
False)
frequent_words = dict(frequent_words)
tf_idf_scores = tfidf.tf_idf(records)
# Pyspark technically ends here - the rest is processed on master node
important_words = tfidf.extract_important_words(tf_idf_scores,
args.num_words, True)
important_words_with_counts = add_word_counts(important_words,
frequent_words)
synset_dict = generate_syn_set(important_words_with_counts.items())
print_syn_set(synset_dict)
