Sklearn cosine_相似性在python中将一维数组转换为二维数组

我正在学习自然语言处理，并在Python中使用nltk模块和scikit学习模块。在编写自己的代码之前，我想先看看现有代码是如何工作的。所以我在网上寻找基于这些库构建的聊天机器人，在github上找到了一个。 我下载了一个聊天机器人的github代码，它使用Scikit学习和nltk模块。这就是代码

from __future__ import division
import numpy as np
import pandas as pd
import sys
import nltk
import pyprind
from nltk.corpus import wordnet as wn
from sklearn.externals import joblib
from sklearn.metrics.pairwise import cosine_similarity 

SYNSETS = joblib.load('blobs/SYNSETS.pkl')
TAGS_HASH = joblib.load('blobs/TAGS_HASH.pkl')
data = pd.read_csv('data/friends-final.txt', sep='\t')
triturns = joblib.load('blobs/triturns.pkl')
filtered_triturns = joblib.load('blobs/filtered.pkl')
all_tags = ['CC', 'CD', 'DT', 'EX', 'FW', 'IN', 'JJ', 'JJR', 'JJS', 'LS', 'MD',
       'NN', 'NNS', 'NNP', 'NNPS', 'PDT', 'POS', 'PRP', 'PRP$', 'RB',
       'RBR', 'RBS', 'RP', 'SYM', 'TO', 'UH', 'VB', 'VBD', 'VBG', 'VBN',
       'VBP', 'VBZ', 'WDT', 'WP', 'WP$', 'WRB']

def get_synsets(text):
    if not text in SYNSETS:
        sent = nltk.pos_tag(nltk.word_tokenize(text))
        chunks = nltk.ne_chunk(sent, binary=False)
        s = set()
        def add_synsets(synsets):
            for synset in synsets:
                s.add(synset)
        for c in chunks:
            if hasattr(c, 'node'):
                if c.node == 'PERSON':
                    add_synsets(wn.synsets('person', pos=wn.NOUN))
                elif c.node == 'ORGANIZATION':
                    add_synsets(wn.synsets('organization', pos=wn.NOUN))                
                elif c.node == 'GPE':
                    add_synsets(wn.synsets('place', pos=wn.NOUN))
                elif c.node == 'LOCATION':
                    add_synsets(wn.synsets('location', pos=wn.NOUN))
                elif c.node == 'FACILITY':
                    add_synsets(wn.synsets('facility', pos=wn.NOUN))
                elif c.node == 'GSP':
                    add_synsets(wn.synsets('group', pos=wn.NOUN))                
                else:
                    print c, c.node, c.leaves()
            elif c[1][:2] in ['VB', 'JJ', 'ADV', 'NN']:
                pos = {'VB': wn.VERB, 'NN': wn.NOUN, 'ADV': wn.ADV, 'JJ': wn.ADJ}[c[1][:2]]
                add_synsets(wn.synsets(c[0], pos=pos))
            else:
                add_synsets(wn.synsets(c[0]))
        SYNSETS[text] = set([x.name for x in s])
    return SYNSETS[text]

def sem_sim(s1, s2):
    ss1 = get_synsets(s1)
    ss2 = get_synsets(s2)
    if ss1 == ss2:
        return 1
    return 2*len(ss1.intersection(ss2)) / (len(ss1) + len(ss2))

def cos_sim(s1, s2):
    d = [{}, {}]
    for p in all_tags:
        d[0][p] = d[1][p] = 0
    for i,s in enumerate([s1, s2]):
        if not s in TAGS_HASH:
            TAGS_HASH[s] = nltk.pos_tag(nltk.word_tokenize(s))
        tags = TAGS_HASH[s]
        for t in tags:
            if t[1] in d[i]:
                d[i][t[1]] += 1
    return cosine_similarity([d[0][p] for p in all_tags], [d[1][p] for p in all_tags])[0][0] #ERROR OCCURS HERE

def sim(s1, s2, alpha=0.7):
    return alpha*sem_sim(s1, s2) + (1-alpha)*cos_sim(s1, s2) #ERROR OCCURS HERE

def get_response(msg):
    best_val = 0
    best = None
    bar = pyprind.ProgBar(len(filtered_triturns), monitor=True)
    for t in filtered_triturns:
        question = t[0]
        answer = t[1]
        val = sim(msg, question)
        if (val > best_val):# or (val == best_val and len(answer) < len(msg))):
            best = answer
            best_val = val
        bar.update()
    return best

def main():
    while True:
        msg = raw_input('--> ')
        print get_response(msg)
        sys.stdout.flush()

def filter_triturns(thresh=0.7):
    L = []
    bar = pyprind.ProgBar(len(triturns), monitor=True)
    for i,tt in enumerate(triturns):
        a = data.irow(tt)['line']
        b = data.irow(tt+1)['line']
        c = data.irow(tt+2)['line']
        if sem_sim(a, b) > thresh:
            L.append([a,b])
        if sem_sim(b, c) > thresh:
            L.append([c,b])
        bar.update()
    return L

if __name__ == '__main__':
    main()

该错误发生在

行返回余弦_相似性（[d[0][p]表示所有_标记中的p]，[d[1][p]表示所有_标记中的p]）[0][0]

及其下方的行（在上面的代码中用#标记所述行）

它基本上是要求我给出一个2D数组，而不是我尝试过的一维数组，但这似乎不起作用。我对sklearn还比较陌生，所以我不太清楚应该在什么时候把代码放在方括号内

将一维数组（如[abc]转换为二维数组）遵循过程[[abc]]，但在这种情况下，我应该在哪里进行转换

---

我使用的是python 2.7

您必须在余弦_相似性计算中对其进行更改，只需更改即可

cosine_similarity([d[0][p] for p in all_tags], [d[1][p] for p in all_tags])[0][0]

到

---

另一种可能是使用

.reformate（）

函数

cosine_similarity([d[0][p] for p in all_tags], [d[1][p] for p in all_tags])[0][0]

cosine_similarity([[d[0][p] for p in all_tags]], [[d[1][p] for p in all_tags]])[0][0]