def tokenize(file_data):
file_data=file_data.lower() #lowercasing
file_data = re.sub('[^0-9a-zA-Z]+', ' ', file_data) #ignoring punctuations
file_data = re.sub('(\w)\'(\w)/$1~$2', ' ', file_data) #ignoring punctuations
file_data = re.sub('\'/ ', ' ', file_data) #ignoring punctuations
file_words = file_data.split() #split the file into single words
count = 0
for word in file_words:
file_words[count] = stem(stem(word))
count+=1
return file_words
def getWeightsForQuery(self, query):
'''
retourne les pondérations idf des termes de la requête
paramètres
----------
query : string
requête
renvoie
-------
token_weights: dict of string -> float
dictionnaire associant à chaque terme de la requête
sa pondération idf
'''
df_dict = self.indexer.get_df()
N = self.indexer.N
tokens = query.split(" ")
processed_query = [porter.stem(t.lower()) for t in tokens]
q = list(set(processed_query))
token_weights = dict()
for t in processed_query:
if t in df_dict:
token_weights[t] = math.log((1+N)/(1+df_dict[t]))
else:
token_weights[t] = 0
return token_weights
def replace_word_by_stems(wordList):
stemList = []
for word in wordList:
stemList.append(stem(word))
return set(stemList)
def stem(tokens):
"""
now we could do this in python. We coudl also call the same function that
exists server-side, or even run an embedded javascript interpreter. See
http://groups.google.com/group/mongodb-user/browse_frm/thread/728c4376c3013007/b5ac548f70c8b3ca
"""
return [porter.stem(tok) for tok in tokens]
def getWeightsForStem(self,stem):
stem = porter.stem(stem.lower())
dicoRes=dict()
indexInverse = self.indexObject.getTfsForDoc(stem)
for doc in self.index:
if doc in indexInverse:
dicoRes[doc]=1+math.log(indexInverse[doc])
return dicoRes
def Query_processing(query):
query_tokens = tokenizer.tokenize(query)
query_tokens = [w.casefold() for w in query_tokens]
query_tokens = [w for w in query_tokens if w not in stop_words]
query_tokens = [ps.stem(w) for w in query_tokens]
return query_tokens, []
def e_morphological_feats(self):
D = {}
etymology = retrieve_etymology(self.lemma)
D["e_latin_root"] = has_ancestor_in_lang("lat", etymology) # check wiktionary
D["e_length_dist_lemma_form"] = len(self.word) - len(self.lemma)
stem, steps = porter.stem(self.word)
D["e_length_dist_stem_form"] = len(self.word) - len(stem)
D["e_inflectional_morphemes_count"] = steps
return D
def getTextRepresentation(self,text):
tab=re.findall(r"\w+",text,re.UNICODE)
tab=[i.lower() for i in tab]
ret=Counter(tab)
ret={porter.stem(a):b for (a,b) in ret.items() if a not in self.stopWords}
return ret
def getTextRepresentation(self,text):
tab=re.findall(r"\w+",text,re.UNICODE)
tab = [porter.stem(word) for word in tab if word not in self.stopWords]
ret=Counter(tab)
# Error programs, program gives only 1
#ret={porter.stem(a):b for (a,b) in ret.items() if a not in self.stopWords}
return dict(ret)
def getWeightsForStem(self,stem):
stem = porter.stem(stem.lower())
dicoRes=dict()
indexInverse = self.indexObject.getTfsForDoc(stem)
for doc in self.index:
if doc in indexInverse:
idf=math.log((1+len(self.index))/(1+len(indexInverse)))
dicoRes[doc]=(1+math.log(indexInverse[doc]))*idf
return dicoRes
def parse_query(content, stopwords):
tempbagofwords = content.replace("-", " ").split()
bagofwords=[]
for word in tempbagofwords:
word = word.lower()
if not stopwords.has_key(word):
word = porter.stem(word)
bagofwords.append(word)
return bagofwords
def query(self, querystring):
'''
'''
query_no_puct = cfc_tools.replace_punctuation(querystring)
query_termlist = cfc_tools.remove_multiple_space(query_no_puct).strip().split()
acc = list()
for term in query_termlist:
term_processed = porter.stem(term)
acc.append((term, self[term].doclist))
def getTextRepresentation(self, text):
tab = re.findall(r"\w+", text, re.UNICODE)
tab = [porter.stem(word.lower()) for word in tab]
tab = [word for word in tab if word not in self.stopWords]
ret = Counter(tab)
ret = {a: b for (a, b) in ret.items()}
return ret
def e_morphological_feats(self):
D = {}
etymology = retrieve_etymology(self.lemma)
D["e_latin_root"] = has_ancestor_in_lang("lat",
etymology) # check wiktionary
D["e_length_dist_lemma_form"] = len(self.word) - len(self.lemma)
stem, steps = porter.stem(self.word)
D["e_length_dist_stem_form"] = len(self.word) - len(stem)
D["e_inflectional_morphemes_count"] = steps
return D
def queryPreprocessing(query):
"""
Preprocess d'une query sous forme de string pour renvoyer un tableau de mots
"""
arrayQuery = "".join(c for c in query if c.isalnum() or c.isspace()).split()
#print(arrayQuery)
res=[]
for word in arrayQuery:
word = porter.stem(word.lower()) #On stem les mots car ils sont stemmés dans l'index
res.append(word)
return res
def getTextRepresentation(self,text):
""" Return the stemmed representation of a string.
:param text: string, input text
:return: dictionary of {stem: frequency}
"""
tab=re.findall(r"\w+",text,re.UNICODE)
tab=[word.lower() for word in tab]
tab = [porter.stem(word) for word in tab if word not in self.stopWords]
ret=Counter(tab)
ret={a:b for (a,b) in ret.items()}
return ret
def parse_document_content(content, stopwords):
tempbagofwords = content.replace("-", " ").split()
bag_of_words_dict=dict()
global number_of_terms
for word in tempbagofwords:
word = word.lower()
if word not in stopwords:
number_of_terms += 1
word = porter.stem(word)
if bag_of_words_dict.has_key(word):
bag_of_words_dict[word] += 1
else:
bag_of_words_dict[word] = 1
return bag_of_words_dict
def Stemming():
for Tokenobj in TokenPages:
Tokenobj.title = [ps.stem(w) for w in Tokenobj.title]
Tokenobj.infobox = [ps.stem(w) for w in Tokenobj.infobox]
Tokenobj.category = [ps.stem(w) for w in Tokenobj.category]
Tokenobj.links = [ps.stem(w) for w in Tokenobj.links]
Tokenobj.ref = [ps.stem(w) for w in Tokenobj.ref]
Tokenobj.body = [ps.stem(w) for w in Tokenobj.body]
def evaluation_modele_vectoriel(D,qry):
qry = [porter.stem(mot) for mot in qry.split()]
ind = index(D)
score = dict()
for doc in ind[0]:
score[doc]=0
for mot in qry:
for docu in ind[1][mot]:
score[int(docu)]+=int(ind[1][mot][docu])
return score
def evaluation_modele_binaire(D,qry):
ind = index(D)[0]
score = dict()
for i in ind:
score[i]=0
for j in qry.split():
if porter.stem(j) in ind[i]:
score[i]+=1
if len(qry.split())==score[i]:
score[i] = 1
else:
score[i] = 0
return score
def input_query():
_result = []
qq = []
q = raw_input("Please Enter Your Query:" + '\n')
x = int(raw_input("pleas enter how many document id you want?" + '\n'))
q = re.findall(r"[\w']+", q)
for i in range(0, len(q)):
q[i] = q[i].lower()
q[i] = porter.stem(q[i])
if (q[i] in list_stop):
continue
qq.append(q[i])
#print qq #################
_result = score_phrase(qq, x)
user(_result)
def create_index_doc(nb):
global allwords
# Reading document and conversion to lower case
words = re.findall(r'\w+', open('doc' + str(nb) + '.txt').read().lower())
# Reading stop words
stop_words = re.findall(r'\w+', open('stop_words.txt').read().lower())
# Deleting stop words + normalization
words = [porter.stem(word) for word in words if word not in stop_words]
#Saving all the words of all the documents
allwords = set(list(allwords) + words)
# Creation of dict
return dict(collections.Counter(words))
def ParseRawData(data,stopWordFile):
'''do tokenize, stemming, remove stop word
return the parsed file name
'''
logging.info('begin parse raw data')
parsedDatafile = '/parsedPatent.txt'
writefile = open(directory + parsedDatafile,'w')
# open stop word file
# stopwordlist = [word for word in open(stopWordFile).read().split(',')]
for line in open(data,'r'):
stemmed_line =[stem(word) for word in cleanTokenizeText(line)]
# nostop_word_list = [word for word in stemmed_line if word not in stopwordlist]
writefile.write(' '.join(stemmed_line)+'\n')
writefile.close()
logging.info('finished parsing, store the parsed file into ./TFIDF_LDA/parsedPatent.txt')
return parsedDatafile
def create_black_list():
fr=None
tmp=[]
fr=open("balcklist_word.txt",'rb')
#s=[]
for line in fr:
line= re.findall(r"[\w']+",line)
for i in range(0,len(line)):
line[i]=line[i].lower()
line[i] = porter.stem(line[i])
if(line[i] in list_stop):
continue
black_list.append(line[i])
tmp.append(line[i])
black_list_s.append(tmp)
tmp=[]
blac_list=list(OrderedDict.fromkeys(black_list))
def Pre_Processing(Abstract):
# case folding
Abstract = Abstract.lower()
#split into words
tokenizer = RegexpTokenizer(r'[a-zA-Z0-9]+')
words_tokens = tokenizer.tokenize(Abstract)
#stop word removal
stop_words = set(stopwords.words('english'))
words_tokens = [token for token in words_tokens if token not in stop_words]
#stemming
words_tokens = [pt.stem(token) for token in words_tokens]
return words_tokens
def create_black_list():
fr = None
tmp = []
fr = open("balcklist_word.txt", 'rb')
#s=[]
for line in fr:
line = re.findall(r"[\w']+", line)
for i in range(0, len(line)):
line[i] = line[i].lower()
line[i] = porter.stem(line[i])
if (line[i] in list_stop):
continue
black_list.append(line[i])
tmp.append(line[i])
black_list_s.append(tmp)
tmp = []
blac_list = list(OrderedDict.fromkeys(black_list))
def ParseRawData(data, stopWordFile):
'''do tokenize, stemming, remove stop word
return the parsed file name
'''
logging.info('begin parse raw data')
parsedDatafile = '/parsedPatent.txt'
writefile = open(directory + parsedDatafile, 'w')
# open stop word file
# stopwordlist = [word for word in open(stopWordFile).read().split(',')]
for line in open(data, 'r'):
stemmed_line = [stem(word) for word in cleanTokenizeText(line)]
# nostop_word_list = [word for word in stemmed_line if word not in stopwordlist]
writefile.write(' '.join(stemmed_line) + '\n')
writefile.close()
logging.info(
'finished parsing, store the parsed file into ./TFIDF_LDA/parsedPatent.txt'
)
return parsedDatafile
def Field_Query_Processing(query):
query_tokens = []
fields = []
tokens = query.split(',')
for token in tokens:
field_tokens = token.split(':')
field = field_tokens[0].strip()
value_tokens = field_tokens[1].split(" ")
for val in value_tokens:
fields.append(field)
query_tokens.append(val.strip())
query_tokens = [w.casefold() for w in query_tokens]
query_tokens = [w for w in query_tokens if w not in stop_words]
query_tokens = [ps.stem(w) for w in query_tokens]
return query_tokens, fields
def doQuery(query, index, k):
q = query.split()
#tokenize query
j = 0
while(j < len(q)):
token = tokenize(q[j])
if(token == ''):
del q[j]
j -= 1
else:
q[j] = porter.stem(token)
j += 1
#tier1sim is list of tuples of the form (docID, similarity)
#similarity is the similarity between document number docID and query q.
tier1Sim = []
i = 0
while(i < index.numDocuments):
tier1Sim.append((i + 1, index.computeSimilarity(q, i + 1, index.tier1)))
i += 1
sortedSim = sorted(tier1Sim, key = lambda tup: tup[1], reverse = True)
sortedSim = sortedSim[0:k]
simTier2 = []
i = 0
while(i < len(sortedSim)):
docID = sortedSim[i][0]
simTier2.append((docID, index.computeSimilarity(q, docID, index.tier2)))
i += 1
resultListTups = sorted(simTier2, key = lambda tup: tup[1], reverse = True)
resultList = []
i = 0
while(i < len(resultListTups)):
resultList.append(resultListTups[i][0])
i += 1
return resultList
def getTextRepresentation(self, text):
"""
Permet d'obtenir une représentation d'un texte, On va connaitre le nombre de tous les mots présents dans
le texte, mot vide exclu
:type text: String
:param text: Le texte dont on veux avoir la representation
:return: un dictionnaire qui reprensente texte, c'est a dire que les mots sont lemmatisé et compté par occurence
{mot1 : n1, mot2 : n2, ...}
"""
mots = re.findall(
r"\w+", text,
re.UNICODE) # On recupére une liste tous les mots du texte
mots = [i.lower() for i in mots] # On met tous les mots en minuscule
compte_mots = Counter(mots)
resultat = {
porter.stem(a): b
for (a, b) in compte_mots.items() if a not in self.stopWords
}
return resultat
def read_doc(position):
_str = str()
with open('DOCUMENT.txt', 'rb') as f:
f.seek(position)
line = f.readline()
end = line.find("</BODY>")
while (end == -1):
_str = _str + line
line = f.readline()
end = line.find("</BODY>")
_str = _str + line[:end]
f.close()
_string = []
_str = re.findall(r"[\w']+", _str)
#print len(_str)
for i in range(0, len(_str)):
_str[i] = _str[i].lower()
_str[i] = porter.stem(_str[i])
if (_str[i] in list_stop):
continue
_string.append(_str[i])
return _string
def getWeightsForQuery(self, query):
'''
retourne les pondérations 0-1 pour les termes de la requête
paramètres
----------
query : string
requête
renvoie
-------
token_weights: dict of string -> int
dictionnaire associant à chaque terme de la requête
la pondération 1
'''
query_lower = query.lower()
tokens = list(set(query_lower.split(" ")))
token_weights = dict()
for t in tokens:
token_weights[porter.stem(t)] = 1
return token_weights
def read_doc(position):
_str=str()
with open('DOCUMENT.txt','rb') as f:
f.seek(position)
line=f.readline()
end=line.find("</BODY>")
while(end==-1):
_str = _str + line
line=f.readline()
end=line.find("</BODY>")
_str = _str + line[:end]
f.close()
_string =[]
_str = re.findall(r"[\w']+",_str)
#print len(_str)
for i in range(0,len(_str)):
_str[i]= _str[i].lower()
_str[i] = porter.stem(_str[i])
if(_str[i] in list_stop):
continue
_string.append(_str[i])
return _string
def getWeightsForQuery(self, query):
'''
retourne les pondérations tf des termes de la requête
paramètres
----------
query : string
requête
renvoie
-------
token_weights: dict of string -> int
dictionnaire associant à chaque terme de la requête
sa pondération tf
'''
tokens = query.split(" ")
processed_query = [porter.stem(t.lower()) for t in tokens]
c = collections.Counter(processed_query)
tokens = np.array(list(c.keys()))
token_counts = np.array(list(c.values()))
token_weights = dict()
for i in range(len(tokens)):
token_weights[tokens[i]] = token_counts[i]
return token_weights
def processPhrase(query, negFlag):
# global docList
global docID
global inv_index
global rank
global STOPWORDS
global phrase_pos
word_docs={}
phrase_words=query.split() #split the phrase into single words
deletionList=[]
for find_stop in phrase_words: #remove stopwords from the phrase
if find_stop in STOPWORDS:
deletionList.append(find_stop)
for find_stop in deletionList:
phrase_words.remove(find_stop)
count=0
for word in phrase_words:
phrase_words[count]=stem(stem(word))
count+=1
phrase_length=len(phrase_words)
for word in phrase_words:
word=stem(stem(word))
if word not in inv_index:
return []
if word in inv_index:
for docs in inv_index[word]:
if word not in word_docs:
word_docs[word]=[docs[0]]
else:
word_docs[word].append(docs[0])
combinedList=[] #stores documents in which all words of the phrase are present
actualList=[] #stores documents in which all words are present at adjacent positions
count=0
for docs in word_docs.values(): #forming the combined list
if count == 0:
combinedList=docs
else:
combinedList=list(set(combinedList) & set(docs))
count+=1
for doc in combinedList: #processing docs in the combined list for adjacent positions of words
phrase_count=0
filename=docID[doc]
words=getDocWords(filename)
word=phrase_words[0]
postings=inv_index[word]
for entry in postings: #takes a word in the phrase and checks for adajacent words in the document
if entry[0] != doc:
continue
phrase_start=0
for position in entry[1]:
check="true"
for i in range(0,phrase_length): #iterates over phrase length to check adjacent positions
if position+i<len(words):
if stem(stem(words[position+i])) != phrase_words[i]:
check="false"
break
elif position+i<len(words) and stem(stem(words[position+i])) == phrase_words[i]:
continue
else:
check="false"
if check == "true":
if doc not in actualList:
actualList.append(doc)
phrase_start=position
phrase_count+=1
if phrase_count != 0: #negation flag is to handle negation of phrases
if negFlag == "false":
if not doc in phrase_pos: #rank documents other than the phrase for negation queries
phrase_pos[doc]=phrase_start #position of the phrase for snippets
if doc not in rank:
rank[doc]=phrase_count*len(phrase_words)
else:
rank[doc]+=(phrase_count*len(phrase_words))
# if phrase_count==0:
# combinedList.remove(doc)
#the result is the intersection of the real list and combined list
combinedList=intersect(actualList,combinedList)
return combinedList
def main_dictionary():
m=[]
current_milli_time = lambda: int(round(time.time() * 1000))
num_doc=[]
doc_id=1
b_start = 0
b_end = 0
tmp=str()
total_temp_files=0
#num_line=0
last_tick=current_milli_time()
print 'reading document file...'
total_records=0
fr = open('DOCUMENT.txt','r')
ss=fr.tell()
num_line=len(fr.readlines())
fr.seek(ss)
position=ss
line=fr.readline()
_num=0
while(_num<num_line):
if (b_start==0):
start=line.find("<BODY>")
if (start > -1):
tmp = tmp + line[start+6:]
pp=position
b_start = 1
if (b_start==1):
end=line.find("</BODY>")
if(end == -1):
tmp = tmp + line
if (end != -1):
tmp = tmp + line[:end]
b_start = 0
b_end =1
if(b_end == 1):
tmp = re.findall(r"[\w']+",tmp)
nd=0
for i in range(0,len(tmp)):
tmp[i]=tmp[i].lower()
tmp[i] = porter.stem(tmp[i])
if(tmp[i] in list_stop):
continue
m.append((tmp[i],[(doc_id,i)]))
nd+=1
num_doc.append((pp,nd))
b_end=0
tmp=str()
doc_id = doc_id + 1
if sys.getsizeof(m)>4*1024*1024:
total_temp_files=total_temp_files+1
fname='partial-temp-'+str(total_temp_files)+'.dat'
print 'saving temporary file ',fname,'...'
m=sort_data(m)
write_file(fname,m,int(0),len(m))
total_records=total_records+len(m)
m=[]
position=fr.tell()
line=fr.readline()
_num+=1
total_temp_files=total_temp_files+1
fname='partial-temp-'+str(total_temp_files)+'.dat'
print 'saving temporary file ',fname,'...'
m=sort_data(m)
write_file(fname,m,int(0),len(m))
total_records=total_records+len(m)
m=[]
now_tick=current_milli_time()
print 'read ',total_records,' records in ',(now_tick-last_tick),' ms.',' and saved ',total_temp_files,' temp files...'
last_tick=now_tick
if total_temp_files==1:
os.rename('partial-temp-1.dat','inverted-index.dat')
else:
merge_files('partial-temp-1.dat','partial-temp-2.dat','merge-temp.dat')
os.remove('partial-temp-1.dat')
os.remove('partial-temp-2.dat')
for i in range(3,total_temp_files+1):
os.rename('merge-temp.dat','merge-temp-2.dat')
print 'merging partial-temp-',str(i),'.dat...'
merge_files('merge-temp-2.dat','partial-temp-'+str(i)+'.dat','merge-temp.dat')
os.remove('merge-temp-2.dat')
os.remove('partial-temp-'+str(i)+'.dat')
os.rename('merge-temp.dat','inverted-index.dat')
now_tick=current_milli_time()
print 'merged inverted-index.dat in ',(now_tick-last_tick),' ms.'
last_tick=now_tick
print 'done!'
unmerged_files()
num_document(num_doc)
print 'done'
# -*- coding: utf-8 -*-
"""
common.py: provides common functions for 'text' module.
"""
import porter
import re
re_sentence = re.compile(r'[^?!.;]+[?!.;]?\n?')
re_word = re.compile(r"[\w]+")
strip_html = lambda s: re.sub(r'</?[^<>]*/?>', ' ', s)
get_stem = lambda w: porter.stem(w.lower())
def remove_camel_case(text):
text = re.sub(r'([a-z])([A-Z])', '\\1 \\2', text)
text = re.sub(r'([0-9])([a-zA-Z])', '\\1 \\2', text)
text = re.sub(r'([a-zA-Z])([0-9])', '\\1 \\2', text)
return text
def is_bad(word):
return re.match('^[0-9]+', word) is not None
def iter_words(text):
sentences = re.finditer(re_sentence, text)
def extract_terms(text):
term_list = [stem(term)
for term in splitter.split(text.lower())
if ((len(term)>1) and (term not in STOP_WORDS))]
return list(frozenset(term_list))
def countWord(doc):
return dict(Counter([porter.stem(i.lower()) for i in doc.split() if i.lower() not in motvide]))
def counter(phrase):
porter_stemer = textRepresenter.PorterStemmer()
#ou textRepresenter.getTextRepresentation
l = (porter.stem(w.lower()) for w in phrase.split(" ") if w != "")
l = (w for w in l if w not in porter_stemer.stopWords)
return dict(collections.Counter(l).items())
def tokenize(word):
regex = re.compile('[^a-zA-Z]+')
w = regex.sub('', word)
w = w.lower()
w = porter.stem(w)
return w
def main():
global docID
global inv_index
global path
global phraseList
global query_result
global docList
global negList
global tf
global df
global allDocs
global rank
global phrase_pos
global phraseList
global snippetDone
while(True):
loadIndex()
path = inv_index['cranfield_corpus_path']
loadDocID()
query=raw_input("\nEnter your query or a special command (tf/df/freq/title/doc/author/similar) (Enter goodbye to exit)\n")
start = time.time()
makeAllDocList()
#convert entire query to lower case
query = query.lower()
if query == 'goodbye':
print "Thank you for using this information retrieval system.."
break
#shlex is used to split pharses and words separately
query_words=shlex.split(query)
count = 0
#stemming of the query
for word in query_words:
if " " not in word:
query_words[count] = stem(stem(word))
count+=1
phraseList=[]
negPhraseList=[]
#process special query 'similar'
if query_words[0] == 'similar':
flag="false"
for word in inv_index.keys():
dist = levenshtein(query_words[1],word)
if dist==1:
flag="true"
print word
if flag == "false":
print "No similar words found"
query_result={} #stores the result of the query for further processing
docList=[]
df={}
negList=[]
tf={}
allDocs=[]
rank={}
phrase_pos={}
phraseList=[]
snippetDone=[]
continue
#process special query tf/df/freq/doc/title/author
if query_words[0] == 'tf' or query_words[0] == 'df' or query_words[0] == 'freq' or query_words[0] == 'doc' or query_words[0] == 'titl' or query_words[0] == 'author':
negFlag="false"
specialQuery(query_words)
query_result={} #stores the result of the query for further processing
docList=[]
df={}
negList=[]
tf={}
allDocs=[]
rank={}
phrase_pos={}
phraseList=[]
snippetDone=[]
continue
#process each word in the query
for word in query_words:
#single word
if word in inv_index and word[0] != '!':
postings=inv_index[word]
docListing(word, postings)
#single word with negation
elif word[0]=='!' and word[1:] in inv_index and " " not in word:
postings=inv_index[word[1:]]
negativeListing(word[1:], postings)
#phrase with negation
elif " " in word and word[0]=='!':
negPhraseList=negPhraseList+processPhrase(word[1:], "true")
#phrase without negation
elif " " in word:
phraseList=phraseList+processPhrase(word, "false")
#take set difference with universal set for phrase negation
if len(negPhraseList) != 0:
negPhraseList=difference(allDocs,negPhraseList)
for doc in negPhraseList:
if doc not in rank:
rank[doc]=1
else:
rank[doc]+=1
#set operations on results
docSet=posUnion() # Set A
docSet=list(set(docSet) | set(phraseList)) # A = A U B
negSet=negUnion() #Set C
negSet=list(set(negSet) | set(negPhraseList)) #C = C U D
resultSet=union(docSet,negSet) # RESULT = A U C = (A U B) U (C U D)
#Display snippets if results have been found
if len(resultSet)!=0:
displaySnippets(resultSet, query_words)
else:
print "No match found!"
print "Number of results:", len(resultSet)
# print resultSet
end = time.time()
print "\nTook", (end-start), "seconds to give the results"
#Re-initilaize global variables for the next query
query_result={} #stores the result of the query for further processing
docList=[]
df={}
negList=[]
tf={}
allDocs=[]
rank={}
phrase_pos={}
phraseList=[]
snippetDone=[]
def textRank(df, docNum, abstractLen):
#Sentence tokenizer
s = str(df[docNum])
sentences = sent_tokenize(str(s))
bagOfWords = {}
i = 0
#For each sentence, clean
for sentence in sentences:
tokens = tokenizer.tokenize(sentence)
finalTokens = {}
for token in tokens:
token = token.casefold()
#Bag of Words
if token not in stopWords and (token.isalpha()
or token.isnumeric()):
token = porter.stem(token)
finalTokens[token] = 1
if len(finalTokens) > 0:
bagOfWords[i] = finalTokens
i += 1
#Matrix N*N (N = no. of sentences)
#Jaccard similarity for each sentence with each sentence (i,j)
lenSecSent = 0
i = 0
j = 0
similarityMatrix = np.zeros((len(sentences), len(sentences)))
for sentence in bagOfWords:
BOW1 = bagOfWords[sentence]
for secondSentence in bagOfWords:
if sentence != secondSentence:
BOW2 = bagOfWords[secondSentence]
lenSecSent = len(BOW2)
j = 0
for token in BOW1:
if token in BOW2:
j += 1
similarityMatrix[sentence][secondSentence] = j / (len(BOW1) +
lenSecSent - j)
similarityMatrix[secondSentence][sentence] = j / (len(BOW1) +
lenSecSent - j)
#Page rank function
similarityMatrix = scipy.sparse.csr_matrix(similarityMatrix)
nx_graph = nx.from_scipy_sparse_matrix(similarityMatrix)
scores = pagerank(nx_graph, max_iter=1000)
#Sorting for top documents
rankedSentences = sorted(((scores[i], s) for i, s in enumerate(sentences)),
reverse=True)
abstract = ""
i = 0
for s in rankedSentences:
if i < abstractLen:
# abstract.append(rankedSentences[i][1])
abstract += rankedSentences[i][1] + " "
i += 1
return abstract
def main_dictionary():
m = []
current_milli_time = lambda: int(round(time.time() * 1000))
num_doc = []
doc_id = 1
b_start = 0
b_end = 0
tmp = str()
total_temp_files = 0
#num_line=0
last_tick = current_milli_time()
print 'reading document file...'
total_records = 0
fr = open('DOCUMENT.txt', 'r')
ss = fr.tell()
num_line = len(fr.readlines())
fr.seek(ss)
position = ss
line = fr.readline()
_num = 0
while (_num < num_line):
if (b_start == 0):
start = line.find("<BODY>")
if (start > -1):
tmp = tmp + line[start + 6:]
pp = position
b_start = 1
if (b_start == 1):
end = line.find("</BODY>")
if (end == -1):
tmp = tmp + line
if (end != -1):
tmp = tmp + line[:end]
b_start = 0
b_end = 1
if (b_end == 1):
tmp = re.findall(r"[\w']+", tmp)
nd = 0
for i in range(0, len(tmp)):
tmp[i] = tmp[i].lower()
tmp[i] = porter.stem(tmp[i])
if (tmp[i] in list_stop):
continue
m.append((tmp[i], [(doc_id, i)]))
nd += 1
num_doc.append((pp, nd))
b_end = 0
tmp = str()
doc_id = doc_id + 1
if sys.getsizeof(m) > 4 * 1024 * 1024:
total_temp_files = total_temp_files + 1
fname = 'partial-temp-' + str(total_temp_files) + '.dat'
print 'saving temporary file ', fname, '...'
m = sort_data(m)
write_file(fname, m, int(0), len(m))
total_records = total_records + len(m)
m = []
position = fr.tell()
line = fr.readline()
_num += 1
total_temp_files = total_temp_files + 1
fname = 'partial-temp-' + str(total_temp_files) + '.dat'
print 'saving temporary file ', fname, '...'
m = sort_data(m)
write_file(fname, m, int(0), len(m))
total_records = total_records + len(m)
m = []
now_tick = current_milli_time()
print 'read ', total_records, ' records in ', (
now_tick -
last_tick), ' ms.', ' and saved ', total_temp_files, ' temp files...'
last_tick = now_tick
if total_temp_files == 1:
os.rename('partial-temp-1.dat', 'inverted-index.dat')
else:
merge_files('partial-temp-1.dat', 'partial-temp-2.dat',
'merge-temp.dat')
os.remove('partial-temp-1.dat')
os.remove('partial-temp-2.dat')
for i in range(3, total_temp_files + 1):
os.rename('merge-temp.dat', 'merge-temp-2.dat')
print 'merging partial-temp-', str(i), '.dat...'
merge_files('merge-temp-2.dat', 'partial-temp-' + str(i) + '.dat',
'merge-temp.dat')
os.remove('merge-temp-2.dat')
os.remove('partial-temp-' + str(i) + '.dat')
os.rename('merge-temp.dat', 'inverted-index.dat')
now_tick = current_milli_time()
print 'merged inverted-index.dat in ', (now_tick - last_tick), ' ms.'
last_tick = now_tick
print 'done!'
unmerged_files()
num_document(num_doc)
print 'done'
def apply_stem(termslist):
'''
'''
for i, term in enumerate(termslist):
termslist[i] = porter.stem(term)
