문자열에 대한 문자열 목록 정렬에 대한 색인

Question

문자열 목록이 큰 문자열에 가장 잘 맞도록 색인을 제공하는 함수가 필요합니다. 예를 들어

:

text = 'Kir4.3 is a inwardly-rectifying potassium channel. Dextran-sulfate is useful in glucose-mediated channels.' 

문자열 목록 : : 문자열을 감안할 때

tok = ['Kir4.3', 'is', 'a', 'inwardly-rectifying', 'potassium', 'channel','.', 'Dextran-sulfate', 'is', 'useful' ,'in', 'glucose','-', 'mediated', 'channels','.'] 

이 기능은 산출하기 위해 만들 수 있습니다

indices = [7, 10, 12, 32, 42, 49, 51, 67, 70, 77, 80, 87, 88, 97, 105] 

---

설명 된대로 기능이 작동하는지

가

from re import split 
from numpy import vstack, zeros 
import numpy as np 

# I need a function which takes a string and the tokenized list 
# and returns the indices for which the tokens were split at 
def index_of_split(text_str, list_of_strings): 
    #????? 
    return indices 

# The text string, string token list, and character binary annotations 
# are all given 
text = 'Kir4.3 is a inwardly-rectifying potassium channel. Dextran-sulfate is useful in glucose-mediated channels.' 
tok = ['Kir4.3', 'is', 'a', 'inwardly-rectifying', 'potassium', 'channel','.', 'Dextran-sulfate', 'is', 'useful' ,'in', 'glucose','-', 'mediated', 'channels','.'] 
# (This binary array labels the following terms ['Kir4.3', 'Dextran-sulfate', 'glucose']) 
bin_ann = [1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0] 

# Here we would apply our function 
indices = index_of_split(text, tok) 
# This list is the desired output 
#indices = [7, 10, 12, 32, 42, 49, 51, 67, 70, 77, 80, 87, 88, 97, 105] 

# We could now split the binary array based on these indices 
bin_ann_toked = np.split(bin_ann, indices) 
# and combine with the tokenized list 
tokenized_strings = np.vstack((tok, bin_ann_toked)).T 

# Then we can remove the trailing zeros, 
# which are likely caused from spaces, 
# or other non tokenized text 
for i, el in enumerate(tokenized_strings): 
    tokenized_strings[i][1] = el[1][:len(el[0])] 
print(tokenized_strings) 

이 가 주어진 다음과 같은 출력을 제공 할 것이다 : 여기

---

내가 점을 설명하기 위해 만든 스크립트입니다

[['Kir4.3' array([1, 1, 1, 1, 1, 1])] 
['is' array([0, 0])] 
['a' array([0])] 
['inwardly-rectifying' 
    array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])] 
['potassium' array([0, 0, 0, 0, 0, 0, 0, 0, 0])] 
['channel' array([0, 0, 0, 0, 0, 0, 0])] 
['.' array([0])] 
['Dextran-sulfate' array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1])] 
['is' array([0, 0])] 
['useful' array([0, 0, 0, 0, 0, 0])] 
['in' array([0, 0])] 
['glucose' array([1, 1, 1, 1, 1, 1, 1])] 
['-' array([0])] 
['mediated' array([0, 0, 0, 0, 0, 0, 0, 0])] 
['channels' array([0, 0, 0, 0, 0, 0, 0, 0])] 
['.' array([0])]] 

Answer 1

text = 'Kir4.3 is a inwardly-rectifying potassium channel. Dextran-sulfate is useful in glucose-mediated channels.' 

tok = ['Kir4.3', 'is', 'a', 'inwardly-rectifying', 'potassium', 'channel','.', 'Dextran-sulfate', 'is', 'useful' ,'in', 'glucose','-', 'mediated', 'channels','.'] 


ind = [0] 
for i,substring in enumerate(tok): 
    ind.append(text.find(substring,ind[i],len(text))) 

print ind[2:] 

의 결과는

[7, 10, 12, 32, 42, 49, 51, 67, 70, 77, 80, 87, 88, 97, 105] 

Answer 2

다음은 무차별 방식입니다. 모든 단어 일치를 찾아서 모든 조합에 점수를 매기는 점수를 얻습니다.

import numpy as np 
from scipy import signal 

def pen(l, r): 
    return (r-l)*(1-4*(l>r)) 

class template: 
    def __init__(self, template): 
     self.template = np.frombuffer(template.encode('utf32'), offset=4, 
             dtype=np.int32) 
     self.normalise = self.template*self.template 
    def match(self, other): 
     other = np.frombuffer(other.encode('utf32'), offset=4, dtype=np.int32)[::-1] 
     m = signal.convolve(self.template, other, 'valid') 
     t = signal.convolve(self.normalise, np.ones_like(other), 'valid') 
     delta = np.absolute(m - t) 
     md = min(delta) 
     return np.where(delta == md)[0], md 
    def brute(self, tok): 
     ms, md = self.match(tok[0]) 
     matches = [[-md, (tok[0], s, s+len(tok[0]))] for s in ms] 
     for t in tok[1:]: 
      ms, md = self.match(t) 
      matches = [[mo[0] - md - pen(mo[-1][-1], mn)] + mo[1:] 
         + [(t, mn, mn + len(t))] for mn in ms for mo in matches] 
     return sorted(matches, key=lambda x: x[0]) 
#   for t in tok[1:]: 
#    ms, md = self.match(t) 
#    matches = [[mo[0] - md] + mo[1:] 
#       + [(t, mn, mn + len(t))] for mn in ms for mo in matches 
#       if mo[-1][-1] <= mn] 
#   return sorted(matches, key=lambda x: x[0]) 

text = 'Kir4.3 is a inwardly-rectifying potassium channel. Dextran-sulfate is useful in glucose-mediated channels.' 
tok = ['Kir4.3', 'is', 'a', 'inwardly-rectifying', 'potassium', 'channel','.', 'Dextran-sulfate', 'is', 'useful' ,'in', 'glucose','-', 'mediated', 'channels','.'] 
tx = template(text) 
matches = tx.brute(tok) 
print(matches[-1]) 

# [-11, ('Kir4.3', 0, 6), ('is', 7, 9), ('a', 10, 11), ('inwardly-rectifying', 12, 31), ('potassium', 32, 41), ('channel', 42, 49), ('.', 49, 50), ('Dextran-sulfate', 51, 66), ('is', 67, 69), ('useful', 70, 76), ('in', 77, 79), ('glucose', 80, 87), ('-', 87, 88), ('mediated', 88, 96), ('channels', 97, 105), ('.', 105, 106)]