def remove_similar_words(word_list, master_list):
# Could replace double for-loops with itertools.combinations ?
new_words = []
stemmer = LancasterStemmer()
for i in sorted(set(word_list)):
x = []
ii = re.sub("[^\w\s]","",i)
ii = re.sub("_","",ii)
ii = re.sub("[ ]"," ",ii)
ii = stemmer.stem(ii)
for j in sorted(master_list):
jj = re.sub("[^\w\s]","",j.decode('UTF-8'))
jj = re.sub("_","",jj)
jj = re.sub("[ ]"," ",jj)
if i[:3]==j[:3]:
jj = stemmer.stem(jj)
if ii == jj:
sim = nltk.edit_distance(i,j)
if sim > 1:
x.append(i)
else:
x.append(i)
x = list(set(x))
new_words+=x
new_words = list(set(new_words))
master_list=list(set(master_list))
new_words = [i for i in new_words if i not in master_list]
master_list.extend(list(set(new_words)))
return new_words, master_list
def lemmatize(s):
lancaster = LancasterStemmer()
words = s.split()
for i in range(0, len(words)):
words[i] = lancaster.stem(words[i])
s = ' '.join(words)
return s
class Tokenizer():
def __init__(self):
self.stemmer = LancasterStemmer()
def __call__(self, text):
return [self.stemmer.stem(token) for token in word_tokenize(text)]
def stem_words(words):
stemmer = LancasterStemmer()
stems_words = []
for word in words:
rstem = stemmer.stem(word)
stems_words.append(rstem + " ")
return stems_words
def stem_text(text):
from nltk.stem import LancasterStemmer
ls = LancasterStemmer()
tokens = tokenize_text(text)
filtered_tokens = [ls.stem(token) for token in tokens]
filtered_text = ' '.join(filtered_tokens)
return filtered_text
def preprocess(raw_string):
stop_words = set(stopwords.words('english'))
res = ''
if raw_string != 'NA':
res = str(raw_string).lower().split()
# Remove stopwords
res = [word for word in res if word not in stop_words]
# Remove punctuation
tokenizer = RegexpTokenizer(r'\w+')
tmp_res = []
for s in res:
tokens = tokenizer.tokenize(s)
tmp_res.extend(tokens)
res = tmp_res
# Stemming
stemmer = LancasterStemmer()
tmp_res = []
for s in res:
if s.isalpha():
tmp_res.append(stemmer.stem(s))
else:
tmp_res.append(s)
res = tmp_res
else:
res = ['NA']
return res
def stemming(tokens):
stemmer = LancasterStemmer()
stems = []
for word in tokens:
stem = stemmer.stem(word)
stems.append(stem)
return stems
def stem_words(words):
stemmer = LancasterStemmer()
stems = []
for word in words:
stem = stemmer.stem(word)
stems.append(stem)
return stems
def lancaster_stemming(lista_doc):
lista = []
stemmer = LancasterStemmer()
for doc in lista_doc:
for word in doc:
lista.append(stemmer.stem(word))
return lista
class Tokenizer(object):
def __init__(self):
self.tok = RegexpTokenizer(r'some_regular_expression')
self.stemmer = LancasterStemmer()
def __call__(self, doc):
return [self.stemmer.stem(token) for token in self.tok.tokenize(doc)]
def strip_sentiment_list(data):
lanc = LancasterStemmer()
list_of_words = list()
for i in data:
new = re.sub(r'http\S+', '', i)
for char in [
'!', '?', ',', '.', '@', '#', '&', '\n', '"', '$', '%', "'",
'(', ')', '*', '-', '+', '/', '^', '[', ']', '”', '_', ':',
';', '|', '{', '}', '~', '€', '£', '“', '1', '2', '3', '4',
'5', '6', '7', '8', '9', '0', '=', '°', 'º', 'ʖ', '…', '⤵',
'↔️'
]:
new = new.replace(char, '')
new = strip_emoji(new)
new = new.lower()
new = new.split(" ")
for j in new:
new_word = lanc.stem(j)
list_of_words.append(new_word)
return strip_stopwords(list_of_words)
class Baseline(object):
def __init__(self):
self.stemmer = LancasterStemmer()
self.stopwords = set([self.stemmer.stem(word) for word in stopwords])
def stem(self, doc):
return [self.stemmer.stem(word) for word in doc]
def doc_similarity(self, s1, s2, pairId=None):
s1 = s1.lower().split()
s2 = s2.lower().split()
s1 = self.stem(s1)
s2 = self.stem(s2)
s1 = set(s1) - self.stopwords
s2 = set(s2) - self.stopwords
return float(len(s1.intersection(s2)))/((len(s1)+len(s2)))
def read_files(categories):
feats = list ()
#porter = PorterStemmer()
lancaster = LancasterStemmer()
print("\n##### Reading files...")
for category in categories:
files = get_filenames_in_folder('Volkskrant/' + category)
num_files=0
for f in files:
data = open('Volkskrant/' + category + '/' + f, 'r', encoding='UTF-8').read()
#data = data.lower()
#data = porter.stem(data)
tokens = word_tokenize(data)
lancaster_list = [lancaster.stem(token) for token in tokens]
#bag = bag_of_words(tokens)
#ww=high_information([(bag, category)], [category])
#bag = bag_of_words(lancaster_list)
#bag = bag_of_non_stopwords(tokens)
bag = bag_of_non_stopwords(lancaster_list)
feats.append((bag, category))
#print len(tokens)
num_files+=1
def getSearchEngineResult(query_dict):
result_dict = {}
ix = index.open_dir("index")
# with ix.searcher(weighting=scoring.BM25F()) as searcher:
with ix.searcher(weighting=scoring.ScoringFunction()) as searcher:
# TODO - Define your own query parser
parser = QueryParser("contents",
schema=ix.schema,
group=OrGroup.factory(0))
stemmizer = LancasterStemmer()
stopWords = set(stopwords.words('english'))
# print(stopWords)
for qid, q in query_dict.items():
table = str.maketrans('\n?.,!', ' ')
q_nomark = q.translate(table)
new_q = ''
for word in q_nomark.split(' '):
if word.lower() not in stopWords:
word_stem = stemmizer.stem(word.lower())
new_q += word_stem + ' '
# print(new_q)
query = parser.parse(new_q.lower())
results = searcher.search(query, limit=None)
# for result in results:
# print(result.fields()['docID'], result.score)
result_dict[qid] = [result.fields()['docID'] for result in results]
return result_dict
def get_word_stems(date, num_subreddits):
import nltk
from nltk.stem import PorterStemmer
from nltk.stem import LancasterStemmer
from nltk.stem import WordNetLemmatizer
table_id = f"top_{num_subreddits}_word_counts"
query = f"""SELECT *
FROM
`{date}.{table_id}`
"""
client = bigquery_client()
job_config = bigquery.QueryJobConfig()
job_config.use_legacy_sql = False
query_job = client.query(query, job_config=job_config)
df = query_job.to_dataframe()
porter = PorterStemmer()
lancaster=LancasterStemmer()
wordnet_lemmatizer = WordNetLemmatizer()
copy = df.copy()
copy['porter']=copy['word'].apply(lambda x: porter.stem(x))
copy['lancaster']=copy['word'].apply(lambda x: lancaster.stem(x))
copy['lemmatised']=copy['word'].apply(lambda x: wordnet_lemmatizer.lemmatize(x))
store_blob(copy, bucket_name=date, blob_name=f"{table_id}_stems.csv")
def preproc(twts):
stop_words = set(stopwords.words('english'))
stab = []
lstem = LancasterStemmer()
REPLACE_NO_SPACE = re.compile("[.;:!\'?,\"()\[\]]")
REPLACE_WITH_SPACE = re.compile("(<br\s*/><br\s*/>)|(\-)|(\/)")
REPLACE_REP = re.compile(r"(\w)\1{2,}")
#abst = [i for sublist in dat for i in sublist]
i = 0
for tweets in twts:
tweets = tweets.lower()
tweets = re.sub(r'[^\w\s]', '', tweets)
tweets = REPLACE_NO_SPACE.sub("", tweets)
tweets = REPLACE_WITH_SPACE.sub(" ", tweets)
tweets = REPLACE_REP.sub("", tweets)
word_tokens = word_tokenize(tweets)
filtered_sentence = [
lstem.stem(w) for w in word_tokens if not w in stop_words
]
stab.append(filtered_sentence)
i += 1
print(i / len(twts))
stab = [remove_num_nword(i) for i in stab]
stab = list(map(lambda x: ' '.join(x), stab))
return stab
def stem_words(words):
stemmer = LancasterStemmer()
stems = []
for word in words:
stem = stemmer.stem(word)
stems.append(stem)
return stems
def process(self):
"""
Apply all the specified preprocessing steps to the tokens from a segment
:return: Tokens, with preprocessing steps applied to them
"""
processed_tokens = self.tokens
if self.stopword_removal:
stop = set(stopwords.words('english'))
processed_tokens = [self.case(token[0]) for token in processed_tokens if token[0] not in stop]
if self.stemming:
if self.stemmer == 'lancaster':
ls = LancasterStemmer()
processed_tokens = [self.case(ls.stem(token[0])) for token in processed_tokens]
elif self.stemmer == 'snowball':
ss = SnowballStemmer('english')
processed_tokens = [self.case(ss.stem(token[0])) for token in processed_tokens]
else:
ps = PorterStemmer()
processed_tokens = [self.case(ps.stem(token[0])) for token in processed_tokens]
if self.lemmatization:
lemma = nltk.wordnet.WordNetLemmatizer()
processed_tokens = [self.case(lemma.lemmatize(token[0], token[1])) for token in processed_tokens]
# no preprocessing method was selected, return only [Bagdad, car,...] instead of [[Bagdad, NNP], [car, NN], [..]..]
if not self.stopword_removal and not self.stemming and not self.lemmatization:
processed_tokens = [self.case(token[0]) for token in processed_tokens]
return processed_tokens
def stem_metn(request):
soz_class = NameForm
cumle_class = TextForm
morf_class = SozForm
porter = PorterStemmer()
lancaster = LancasterStemmer()
k = request.POST.get('metn', '')
alqo = request.POST.get('alqo', '')
txt = k
if alqo == 'Bizim Alqoritm':
txt = metn_oxu(k)
elif alqo == 'Porter Alqoritmi':
txt = porter.stem(txt)
elif alqo == 'Lancaster Alqoritmi':
txt = lancaster.stem(txt)
elif alqo == 'WordNet Alqoritmi':
wordnet_lemmatizer = WordNetLemmatizer()
txt = metn_oxu(wordnet_lemmatizer.stem(k))
return render(request, 'metn.html', {
'form': soz_class,
'cumle': cumle_class,
'morf': morf_class,
'txt': txt
})
def train_bayes(train_list, N_d):
log_prior = {}
V = set()
big_doc = {}
log_likelihood = {}
alpha = 0.5
lc = LancasterStemmer()
tags = ['NN', 'NNS', 'NNP', 'NNPS']
stop_words = set(stopwords.words('english'))
for c in train_list.keys():
# log_prior
for file in train_list[c]:
if c in log_prior:
log_prior[c][0] = log_prior[c][0] + 1 / N_d
log_prior[c][1] = math.log(log_prior[c][0])
else:
log_prior[c] = []
log_prior[c].append(1 / N_d)
log_prior[c].append(math.log(1 / N_d))
# V & big_doc
cur_file = open(file, 'r')
token_file = sent_tokenize(cur_file.read())
for sent in token_file:
token_sent = word_tokenize(sent)
tagged_sent = nltk.pos_tag(token_sent)
for word in tagged_sent:
weight = 1
if word[1] in tags:
weight = 2
stem_word = lc.stem(word[0])
if stem_word not in stop_words:
if stem_word not in V:
V.add(stem_word)
if c in big_doc:
if stem_word in big_doc[c]:
big_doc[c][stem_word] += weight
else:
big_doc[c][stem_word] = weight
else:
big_doc[c] = {}
big_doc[c][stem_word] = weight
cur_file.close()
big_doc_size = {}
for c in train_list.keys():
big_doc_size[c] = sum(big_doc[c].values())
# log_likelihood
for c in train_list.keys():
for w in V:
if w not in big_doc[c]:
log_likelihood[(w, c)] = math.log(alpha / (big_doc_size[c] +
(alpha * len(V))))
else:
log_likelihood[(w, c)] = math.log(
(big_doc[c][w]) / (big_doc_size[c]))
return log_prior, log_likelihood, V
def stemmingLS(self):
from nltk.stem import LancasterStemmer
ls = LancasterStemmer()
temp = copy.deepcopy(self.data)
for i in range(len(temp)):
for j in range(len(temp[i][1])):
temp[i][1][j] = ls.stem(temp[i][1][j])
return temp
def lancaster_stemmer(text):
lancaster = LancasterStemmer()
stemmed_words = []
for word in text.split(" "):
stemmed_words.append(lancaster.stem(word))
return " ".join(stemmed_words)
def token_stemming(x):
'''Uses the LancasterStemmer from nltk to stem tokens in the provided
list of lists `x`'''
stemmer = LancasterStemmer()
x = [[stemmer.stem(w) for w in s] for s in x]
return x
def calculate_uncommon_words_percent(text_file_name):
# read the text from a file
text_file = open(text_file_name)
text = text_file.read()
text_file.close()
# replace punctuation with empty strings
punctuation_marks = [
'.', ',', ':', ';', '(', ')', '!', '?', '[', ']', '$', '"', "'", '’',
'-', '–', '●', '\t', '“', '”', '\n', '0', '1', '2', '3', '4', '5', '6',
'7', '8', '9'
]
for punctuation_mark in punctuation_marks:
text = text.replace(punctuation_mark, ' ')
# chop the text into words, using spaces as word dividers
words_in_text = text.split()
stemmer = LancasterStemmer()
stems_in_text = []
for word in words_in_text:
stem = stemmer.stem(word)
lower_case_stem = stem.lower()
stems_in_text.append(lower_case_stem)
# read the 1000 most common English words from a file
common_words_file = open('most_common_words.txt')
common_words_text = common_words_file.read()
common_words_file.close()
common_words = common_words_text.split()
common_stems = []
for word in common_words:
stem = stemmer.stem(word)
lower_case_stem = stem.lower()
common_stems.append(lower_case_stem)
uncommon_stems = []
for stem in stems_in_text:
if not (stem in common_stems):
uncommon_stems.append(stem)
pct_uncommon_stems = len(uncommon_stems) / len(stems_in_text)
rounded_pct_uncommon_stems = round(pct_uncommon_stems, 3)
return rounded_pct_uncommon_stems
def stem_words(words):
"""Stem words in list of tokenized words"""
stemmer = LancasterStemmer()
stems = []
for word in words:
stem = stemmer.stem(word)
stems.append(stem)
return stems
def get_stems(self):
stemmerlan = LancasterStemmer()
stems = []
for i in self.words:
stems.append(stemmerlan.stem(i))
print("\n提取词干后：")
print(stems)
return stems
def lancaster_stem(tokenize_list):
l = LancasterStemmer()
result = []
for w in tokenize_list:
result.append(l.stem(w))
return result
def stem_words(row):
stemmed_row = ''
lancaster = LancasterStemmer()
for word in row:
word = word.strip()
stemmed_word = lancaster.stem(word)
stemmed_row += stemmed_word + ' '
return stemmed_row
def stem_words(words):
stemmer = LancasterStemmer()
stems_words = []
for word in words:
stem = stemmer.stem(word)
stems_words.append(stem)
stems_words.append(' ') #add space between words
return stems_words
def stemSentence(sentence):
lancaster = LancasterStemmer()
token_words = word_tokenize(sentence)
stem_sentence = []
for word in token_words:
stem_sentence.append(lancaster.stem(word))
stem_sentence.append(" ")
return "".join(stem_sentence)
def stem_words(wrd):
stemmer = LancasterStemmer() # Selects the stemmmer from nltk
stems = [] # List of updated words
for word in wrd:
stem = stemmer.stem(word) # Stems the word
stems.append(stem) # and appends it to the list
return stems
def words_stem(tokens):
"""
convert all words to root of that word
:param tokens:
:return:
"""
from nltk.stem import PorterStemmer, LancasterStemmer
stemmer = LancasterStemmer()
return [stemmer.stem(word) for word in tokens]
def lemmatization(lista_palabras_sp_ssw_scr):
stemmerlan = LancasterStemmer()
lista_palabras_sp_ssw_scr_stem = []
for word in lista_palabras_sp_ssw_scr:
lista_palabras_sp_ssw_scr_stem.append(stemmerlan.stem(word))
tokens = set(lista_palabras_sp_ssw_scr_stem)
#print(tokens)
return tokens
class TFIDF:
def __init__(self):
self.pickle_docs = "tfidf_pickle_docs"
self.pickle_corpus = "tfidf_pickle_corpus"
self.lan = LancasterStemmer()
self.construct()
#print sorted(self.words.iteritems(), key = operator.itemgetter(1), reverse=True)[:20]
def clean(self, word):
'''cleans a word or returns None if it should not be considered'''
word = word.strip(string.punctuation)
word = self.lan.stem(word)
return word
def construct(self):
corpus = {}
# Check to see if we should simply load a pickle
if os.path.isfile(self.pickle_docs):
with open(self.pickle_docs) as docs_file:
current_doclist = pickle.load(docs_file)
if os.listdir('articles/') == current_doclist:
# current article list is the same as pickled article list
# so we want to just load the stored pickled corpus data
with open(self.pickle_corpus) as corpus_file:
self.words = pickle.load(corpus_file)
self.n = len(current_doclist)
return
# If we don't load a pickle, build the corpus from articles/ dir
num_docs = 0.0
for file_name in os.listdir('articles/'):
num_docs += 1
doc = {}
with open("articles/" + file_name) as article:
for line in article:
for word in tokenize(line, "word", return_spans=False):
word = self.clean(word)
doc[word] = 1
for key in doc.keys():
corpus[key] = corpus.get(key, 0) + 1
self.words = corpus
self.n = num_docs
print "Pickling a new TFIDF corpus"
# pickle corpus and document list
with open(self.pickle_docs, "w") as docs_file:
pickle.dump(os.listdir('articles/'), docs_file)
with open(self.pickle_corpus, "w") as corpus_file:
pickle.dump(self.words, corpus_file)
def weight(self, word, count, debug=False):
if debug:
return (word, count, self.words.get(word, 1))
return count * math.log(self.n / self.words.get(word, 1))
def stemWords(self, words): """Stem words in list of tokenized words""" if stemmer == "lancaster": stemmer = LancasterStemmer() elif stemmer == "snowbal": stemmer = SnowballStemmer() elif stemmer == "porter": stemmer = PorterStemmer() stems = [stemmer.stem(word) for word in words] return stems
def stemming(word):
# Use stemmers for removing morphological affixes from words.
Portst = PorterStemmer()
Landst = LancasterStemmer()
Regst = RegexpStemmer('ing|ed')
new = Portst.stem(word)
if new == word:
new = Landst.stem(word)
if new == word:
new = Regst.stem(word)
return new
class StemTokenizer(object):
def __init__(self, stemmer_type='Porter'):
self.stemmer_type = stemmer_type
if self.stemmer_type == 'Porter':
self.stemmer = PorterStemmer()
elif self.stemmer_type == 'Lancaster':
self.stemmer = LancasterStemmer()
else:
raise Exception('Invalid stemmer_type = {0}'.format(stemmer_type))
def __call__(self, doc):
return [self.stemmer.stem(t) for t in word_tokenize(doc)]
def stem(self, input_text):
tokenizer = RegexpTokenizer("\s+", gaps=True)
stemmed_text = []
lemmatizer = WordNetLemmatizer()
stemmer = LancasterStemmer()
text = tokenizer.tokenize(str(input_text))
filtered_text = self.stopword(text)
for word in filtered_text:
if word.isalpha():
stemmed_text.append(stemmer.stem(word).lower())
" ".join(stemmed_text)
return stemmed_text
class LancasterTokenizer(object):
def __init__(self):
self.ls = LancasterStemmer()
self.rx = RegexpTokenizer(r"(?u)\b\w\w+\b")
def isNumber(self, s):
try:
float(s)
return True
except ValueError:
return False
def __call__(self, doc):
return [self.ls.stem(t) for t in self.rx.tokenize(doc) if not self.isNumber(t)]
class Tokenizer():
"""
Tokenizes and stems text using NLTK libraries
"""
def __init__(self):
"""
Constructs a tokenizer object
"""
self.stemmer = LancasterStemmer()
def __call__(self, text):
"""
Tokenizes text
:param text: the text to tokenize
:type text: str or unicode
:return: a list of tokens
:rtype: list of (str or unicode)
"""
return [self.stemmer.stem(token) for token in word_tokenize(text)]
def word_refiner(*args):
Portst = PorterStemmer()
Landst = LancasterStemmer()
Regst = RegexpStemmer('ing|ed|ly|lly')
args = [i for i in args if isinstance(i, unicode)]
for w in map(str, args):
if w in dic1:
yield w
else:
st1 = Portst.stem(w)
if st1 in dic1:
yield st1
else:
st2 = Landst.stem(w)
if st2 in dic1:
yield st2
else:
st3 = Regst.stem(w)
if st3 in dic1:
yield st3
else:
yield w
def preprocess(sentence):
output_list = []
#CASE FOLDING [NOT COMPLETE]
sentence = sentence.lower()
#DATA CLEANING
sentence = sentence.replace('[https://]?[t.co/]?','')
sentence = sentence.replace('@','')
sentence = sentence.replace('[#]?','')
sentence = sentence.replace('[RT]?','')
sentence = sentence.replace(',','')
sentence = sentence.replace('!','')
sentence = sentence.replace('?','')
sentence = sentence.replace('.','')
sentence = sentence.replace('\'','')
sentence = sentence.replace('\"','')
sentence = sentence.replace(':','')
#REMOVE REPEATED CHARS
#sentence = re.sub(r'(\w)\1+', r'\1', sentence)
#TOKENIZE
tt = TweetTokenizer()
temp = tt.tokenize(sentence)
#REMOVE STOP WORDS
stop = stopwords.words('english')
#STEMMING
ls = LancasterStemmer()
newtemp = [eachword for eachword in temp if eachword not in stop]
for eachword in newtemp:
output_list.append(ls.stem(eachword))
return output_list
class LancasterTokenizer(object):
def __init__(self):
self.wnl = LancasterStemmer()
def __call__(self, doc):
return [self.wnl.stem(t) for t in wordpunct_tokenize(doc)]
import nltk
from nltk.stem import LancasterStemmer
stemmerlan=LancasterStemmer()
print(stemmerlan.stem('working'))
print(stemmerlan.stem('happiness'))
Tokens2.append(w)
#fix ascii again, don't know whats happening here!
for w in Tokens2:
for char in w:
if ord(char) > 128:
w.
#stemming
Tokens3 = []
for w in Tokens2:
Tokens3.append(lanStem.stem(w))
return Tokens3
test = clean(tesSum1)
#read files
tesSum1 = teslaSummary1.read()
tesSum5 = teslaSummary5.read()
#stpWrds = set(stopwords.words("english"))
#print stpWrds
sum1SentTok = sentTok(tesSum1)
sum2SentTok = sentTok(tesSum2)
sum1WordTok = wordTok(tesSum1)
sum2WordTok = wordTok(tesSum2)
for w in sum1WordTok:
tryAgain + LancasterStemmer.stem('stemming')
else:
urlUsed.add(url)
the_page = Response.read()
#url parser
soup = BeautifulSoup(the_page,'html.parser')
p=soup.findAll('p')
for ps in p:
psStr = str(ps.get_text().encode('utf-8'))
psStr = re.sub(r'[^a-zA-Z\s\n]',' ',psStr)
if len(psStr)>2:
psStr = nltk.word_tokenize(psStr)
for ss in psStr:
wordLemmatized = wnl.lemmatize(ss)
wordStemed = stem.stem(wordLemmatized)
tempArticle += str(wordStemed).lower()+' '
# print(ps.get_text())
if len(tempArticle) > 800:
r += 1
writer = open('texts/'+str(r),'w')
writer.write(tempArticle)
writer.close
writer = open('labellist3','a')
writer.write(url + '\t' + str(r) + os.linesep)
writer.close
# if hasChinese(ps.get_text()):
# else:
# porter stemmer
from nltk.stem import PorterStemmer
ps = PorterStemmer()
print ps.stem('jumping'), ps.stem('jumps'), ps.stem('jumped')
print ps.stem('lying')
print ps.stem('strange')
# lancaster stemmer
from nltk.stem import LancasterStemmer
ls = LancasterStemmer()
print ls.stem('jumping'), ls.stem('jumps'), ls.stem('jumped')
print ls.stem('lying')
print ls.stem('strange')
# regex stemmer
from nltk.stem import RegexpStemmer
rs = RegexpStemmer('ing$|s$|ed$', min=4)
print rs.stem('jumping'), rs.stem('jumps'), rs.stem('jumped')
print rs.stem('lying')
print rs.stem('strange')
def stemmed(self,word): stemmer = LancasterStemmer() return stemmer.stem(word)
def stem_document(document): from nltk.stem import LancasterStemmer stemmer = LancasterStemmer() return stemmer.stem(document)
# <h2>Stemming Words</h2>
# <p>Stemming is the process of removing <em>affixes</em> from a word to obtain it's root, or <em>stem</em>. For example, the stem of <strong>
# growing</strong> is <strong>grow</strong>. </p>
# <p>Python includes 4 stemming algorithms, 3 of which are demonstrated below. The fourth, <em>Snowball</em> is for non-English languages
# and is not covered here but is in the text </p>
# <codecell>
from nltk.stem import PorterStemmer, LancasterStemmer, RegexpStemmer
porter = PorterStemmer()
lancaster = LancasterStemmer()
reg = RegexpStemmer('ing')
g = 'growing'
print 'Porter yields: ',porter.stem(g)
print 'lancaster yields: ', lancaster.stem(g)
print 'Regexp yields: ', reg.stem(g)
# <markdowncell>
# <p>The output of various words can be different between stemmers:</p>
# <codecell>
g = 'cookery'
print 'Porter yields: ',porter.stem(g)
print 'lancaster yields: ', lancaster.stem(g)
print 'Regexp yields: ', reg.stem(g)
# <markdowncell>
__author__ = "pratap"
# Porter Stemmer
from nltk.stem import PorterStemmer
stemmer = PorterStemmer()
print stemmer.stem("cooking")
print stemmer.stem("cookery")
# Lancaster Stermmer
from nltk.stem import LancasterStemmer
lanc_stemmer = LancasterStemmer()
print lanc_stemmer.stem("cooking")
print lanc_stemmer.stem("cookery")
class TFIDF(object):
def __init__(self, tfidf_file, id2wordFile=None):
self.model = models.TfidfModel.load(tfidf_file)
self.stemmer = LancasterStemmer()
self.stopwords = set([self._preprocess_word(word) for word in stopwords])
#self.stem_model()
print "done"
def _preprocess_word(self, word):
return self.stemmer.stem(word.lower())
#return word.lower()
def stem(self, doc):
return [self.stemmer.stem(word) for word in doc]
def stem_model(self):
print "stemming"
new_id2word = corpora.Dictionary()
# Create a new dicitonary with the stemmed terms and summed document frequencies
for termid, freq in self.model.dfs.iteritems():
stemmed_word = self.stemmer.stem(self.model.id2word[termid])
stemmed_id = None
if stemmed_word in new_id2word.token2id:
stemmed_id = new_id2word.token2id[stemmed_word]
else:
stemmed_id = len(new_id2word.token2id)
new_id2word.token2id[stemmed_word] = stemmed_id
new_id2word.dfs[stemmed_id] = 0
new_id2word.dfs[stemmed_id] += freq # add df from old dicionary
new_id2word.num_docs = self.model.id2word.num_docs
new_id2word.num_nnz = self.model.id2word.num_nnz
new_id2word.num_pos = self.model.id2word.num_pos
self.model.id2word = new_id2word
self.model.dfs = self.model.id2word.dfs
self.model.idfs = precompute_idfs(self.model.wglobal, self.model.dfs, self.model.num_docs)
self.model.save('models/all_lancaster.tfidfmodel')
print len(new_id2word)
print "done stemming"
def restrict_vocab(self, corpus):
vocab = set()
for doc in corpus:
for idx, freq in doc:
vocab.add(idx)
for idx in vocab:
dfs[idx] = self.model.dfs[idx]
idfs[idx] = self.model.idfs[idx]
self.model.dfs = dfs
self.model.idfs = idfs
def to_bow(self, doc):
doc = [self._preprocess_word(word) for word in doc.lower().split() if word not in self.stopwords]
return self.model.id2word.doc2bow(doc)
def doc_similarity(self, s1, s2, pairId=None):
# tfidf1 = self.model[self.to_bow(s1)]
# tfidf2 = self.model[self.to_bow(s2)]
# index = similarities.MatrixSimilarity([tfidf1],num_features=len(self.model.id2word))
# return math.sqrt(index[tfidf2][0])*4. + 1
tfidf1 = self.model[self.to_bow(s1)]
tfidf2 = self.model[self.to_bow(s2)]
common_terms = set(zip(*tfidf1)[0])&set(zip(*tfidf2)[0])
similarity = 0.
tfidf_total = 0.
for term, freq in tfidf1 + tfidf2:
if term in common_terms:
similarity += freq
tfidf_total += freq
val = math.sqrt(similarity/tfidf_total)*5.
if val < 1.: val +=1.
return val
def LancasterTokenizer(s): from nltk import word_tokenize from nltk.stem import LancasterStemmer stemmer = LancasterStemmer() return [stemmer.stem(t) for t in word_tokenize(s)]