NLP实践：计算三篇文章的余弦相似度

import numpy as np
import jieba
import json
import math

class DocumentSimilarity():

	def __init__(self):
		self.pathA = r'合成生物学.txt'
		self.pathB = r'自动化专业.txt'
		self.pathC = r'计算机专业.txt'
		self.mySlot = r'slot.json'
		self.stopWords = ['，','。','《','》','；','：','‘','’','”','“','？','！','的']


	def get_slot(self)->None:

		allStr = ''
		with open(self.pathA,'r',encoding='utf-8') as f_obj:
			allStr += f_obj.read()
		with open(self.pathB,'r',encoding='utf-8') as f_obj:
			allStr += f_obj.read()
		with open(self.pathC,'r',encoding='utf-8') as f_obj:
			allStr += f_obj.read()

		orginalSlot = list(set(jieba.lcut(allStr)))

		slot = [item for item in orginalSlot if item not in self.stopWords]

		with open(self.mySlot,'w') as f_obj:
			f_obj.write(json.dumps(slot))

	def show_slot(self)->None:

		with open(self.mySlot,'r') as f_obj:
			print(json.loads(f_obj.read()))

		input('点击任意键关闭窗口')

	def get_TF(self,slot,path):

		with open(path,'r',encoding='utf-8') as f_obj:
			items = jieba.lcut(f_obj.read())

		TF = list(np.zeros(len(slot),int))

		for item in items:
			if item not in self.stopWords:
				TF[slot.index(item)] += 1

		return TF

	def get_all_TF(self):
		
		with open(self.mySlot,'r') as f_obj:
			slot = json.loads(f_obj.read())

		TF_A = self.get_TF(slot,self.pathA)
		TF_B = self.get_TF(slot,self.pathB)
		TF_C = self.get_TF(slot,self.pathC)

		return TF_A,TF_B,TF_C



	def get_similarity(self,x,y):
		
		x = np.array(x)
		y = np.array(y)

		return np.dot(x,y) / (math.sqrt(sum(pow(x,2))) * math.sqrt(sum(pow(y,2))))

	def show_result(self):
		TF_A,TF_B,TF_C = self.get_all_TF()

		print(self.get_similarity(TF_A,TF_B))
		print(self.get_similarity(TF_B,TF_C))

if __name__ == '__main__':
	ds = DocumentSimilarity()
	ds.show_result()