from sklearn.feature_extraction.text import CountVectorizer
from pymorphy2 import MorphAnalyzer
import numpy as np
import pandas as pd

def finder(corpus, text, number=3):
    def cosine_sim(v1, v2):
        v1 = np.array(v1)
        v2 = np.array(v2)
        res = v1 @ v2
        v1_norm = np.sqrt(np.sum([i**2 for i in v1]))
        v2_norm = np.sqrt(np.sum([i**2 for i in v2]))
        return res / (v1_norm * v2_norm)

    vect = CountVectorizer()
    corpus_t = vect.fit_transform(corpus).toarray()
    text_t = vect.transform([text]).toarray()[0]

    percents = []
    for i in range(len(corpus)):
        percents.append(cosine_sim(corpus_t[i], text_t))

    sort = sorted(percents, reverse=True)
    index = []

    for i in range(number):
        if i == 0:
            index.append(percents.index(sort[i+1]))
        else:
            for j in range(len(index)):
                if percents.index(sort[i+1]) == index[j]:
                    ind = percents.index(sort[i+1])
                    percents.remove(percents[ind])
                    index.append(percents.index(sort[i+1]))
                    break
                else:
                    index.append(percents.index(sort[i+1]))
                    break

    res = [corpus[index[i]] for i in range(number)]

    return res


work = finder(corpus[0:10], corpus[1])
print('Наш текст:\n' + corpus[1] + '\nПохожие')
for i in range(len(work)):
    print(work[i])