因為輿情監測這邊涉及到一些文本相似度的判斷,實作把一類新聞的分類到同一個主新聞下。有點類似baidu相似新聞的搞法。所有抽時間看了些簡單的文本相似度算法。
下面是之前看的萊文斯坦距離算法。大家可以bing一下理論,這裡直接上code。
def levenshtein_distance(first, second):
if len(first) == 0 or len(second) == 0:
return len(first) + len(second)
first_length = len(first) + 1
second_length = len(second) + 1
distance_matrix = [list(range(second_length)) for i in list(range(first_length))] # 初始化矩陣
for i in range(1, first_length):
for j in range(1, second_length):
deletion = distance_matrix[i-1][j] + 1
insertion = distance_matrix[i][j-1] + 1
substitution = distance_matrix[i-1][j-1]
if first[i-1] != second[j-1]:
substitution += 1
distance_matrix[i][j] = min(insertion, deletion, substitution)
return distance_matrix[first_length-1][second_length-1]
if __name__ == '__main__':
print(levenshtein_distance(u"我們不要垃圾消息", u"A垃圾資訊我們不要")) # 運作結果為:2
import Levenshtein
a =r"C:/Users/Administrator/Desktop/a.txt"
b =r'C:/Users/Administrator/Desktop/b.txt'
aa = ""
bb = ""
with open(a,'r') as f:
aa = f.read()
with open (b, 'r') as f1:
bb = f1.read()
print(Levenshtein.distance(a,b))
print(Levenshtein.hamming(a,b))
print(Levenshtein.ratio(aa,bb))
下面的截圖是從網上抄襲過來的。我覺得對于說明這個算法很好。
![](https://img.laitimes.com/img/__Qf2AjLwojIjJCLyojI0JCLiAzNfRHLGZkRGZkRfJ3bs92YsYTMfVmepNHL900VhRHbXp1Mk1mYoR2MMBjVtJWd0ckW65UbM5WOHJWa5kHT20ESjBjUIF2X0hXZ0xCMx81dvRWYoNHLrdEZwZ1Rh5WNXp1bwNjW1ZUba9VZwlHdssmch1mclRXY39CXldWYtlWPzNXZj9mcw1ycz9WL49zZuBnL2YTM4MDOzcTM0ITMwkTMwIzLc52YucWbp5GZzNmLn9Gbi1yZtl2Lc9CX6MHc0RHaiojIsJye.png)