(W_ij, W, word_freq_1, word_freq_2)
| 52 | |
| 53 | |
| 54 | def cal_pmi(W_ij, W, word_freq_1, word_freq_2): |
| 55 | p_i = word_freq_1 / W |
| 56 | p_j = word_freq_2 / W |
| 57 | p_i_j = W_ij / W |
| 58 | pmi = math.log(p_i_j / (p_i * p_j)) |
| 59 | |
| 60 | return pmi |
| 61 | |
| 62 | |
| 63 | def count_pmi(windows_len, word_pair_count, word_window_freq, threshold): |