tf_idf.py

import glob
import math
line=''
s=set()
flist=glob.glob(r'E:\PROGRAMMING\PYTHON\programs\corpus2\*.txt') #get all the files from the d`#open each file >> tokenize the content >> and store it in a set
for fname in flist:         
    tfile=open(fname,"r")
    line=tfile.read() # read the content of file and store in "line"
    tfile.close() # close the file
    s=s.union(set(line.split(' '))) # union of common words

s=sorted(s) # sorts the content alphabetically


i=0
ct=0
tf_line=''          
doc_counts=[]       
for term in s: #takes each term in the set 
    doc_counts.append(0)
    
    for fdoc in flist: # counts the no of times "term" is encountered in each doc
        
        doc=open(fdoc)
        line=doc.read()
        doc.close()
        ct=line.count(str(term)) #counts the no. of times "term" is present in the file
        tf_line+=str(ct)+',' #prints the count in each doc seperated by comma
        if (ct>0):              #counts no of docs in which 
            doc_counts[i]+=1    #this "term" is found
    i+=1
    tf_line=tf_line.strip()+'\n'    

idf=[]  #inverse document frequency      
weights=[]      #weight
total_docs=len(flist)   #total number of documents

i=0

for doc_count in doc_counts:    #takes the 1st doc count
    idf.append(math.log(total_docs/doc_count)) #calculates idf for each "term"
    weights.append(idf[i]*doc_count) #calculate weight of the term
    i+=1



final_line='TERM'+','       
i=1
for f in flist:
    final_line+='D'+str(i)+'  '+','
    i+=1
final_line+=','+'IDF'+','+'TF-IDF\n'


tf_arr=tf_line.split('\n')



i=0
for term in s:
    final_line+=term+','+tf_arr[i]+','+str(round(idf[i],2))+','+' '+str(round(weights[i],2))+','+'\n'
    i+=1

fdoc="E:\\PROGRAMMING\\PYTHON\\IRprax\\tftable.csv"
outfile=open(fdoc,"w")
outfile.write(final_line)
outfile.close()

print "DONE"