Co occurance matrix for tfidf vectorizer for top 2

问题:

i computed tfidf vectorizer for text data and got vectors as (100000,2000) max_feature = 2000.

while i am computing the co occurance matrix by below code.

length = 2000
m = np.zeros([length,length]) # n is the count of all words
def cal_occ(sentence,m):
    for i,word in enumerate(sentence):
    print(i)
    print(word)
    for j in range(max(i-window,0),min(i+window,length)):
        print(j)
        print(sentence[j])
        m[word,sentence[j]]+=1
for sentence in tf_vec:
    cal_occ(sentence, m)

I am getting the following error.

0
(0, 1210)   0.20426932204609685
(0, 191)    0.23516811545499153
(0, 592)    0.2537746177804585
(0, 1927)   0.2896119458034052
(0, 1200)   0.1624114163299802
(0, 1856)   0.24376566018277918
(0, 1325)   0.2789314085220367
(0, 756)    0.15365704375851477
(0, 1130)   0.293489555928974
(0, 346)    0.21231046306681553
(0, 557)    0.2036759579760878
(0, 1036)   0.29666992324872365
(0, 264)    0.36435609585838674
(0, 1701)   0.242619998334931
(0, 1939)   0.33934107208095693
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-96-ad505b6df734> in <module>()
 11             m[word,sentence[j]]+=1
 12 for sentence in tf_vec:
 ---> 13     cal_occ(sentence, m)

 <ipython-input-96-ad505b6df734> in cal_occ(sentence, m)
  9             print(j)
 10             print(sentence[j])
 ---> 11             m[word,sentence[j]]+=1
 12 for sentence in tf_vec:
 13     cal_occ(sentence, m)

IndexError: only integers, slices (:), ellipsis (...), numpy.newaxis (None) and integer or boolean arrays are valid indices

回答1:

You are having the problem most probably here:

for j in range(max(i-window,0),min(i+window,length)):

min function returns length when i+window exceeds the bound, can you try this instead of the line above:

for j in range(max(i-window,0),min(i+window,length-1)):

Hope this helps,

Cheers