Euclidean.py

from sentence_transformers import SentenceTransformer
import re
import numpy as np
model = SentenceTransformer('paraphrase-distilroberta-base-v1')


def cleanText(readData):
    text = re.sub(
        '[-=+,#/\?:^$.@*\"※~&%ㆍ!』\\‘|\(\)\[\]\<\>`\'…》]', '', readData)
    return text


def euclidean(testList, patternIdx):

    for i in range(len(testList)):
        testList[i]["processed"] = cleanText(testList[i]["title"])

    for i in range(len(testList)):
        patternIdx = patternIdx - 1
        sentences = [testList[i]["processed"],
                     testList[patternIdx]["processed"]]
        # (1) 이용하여 임베딩
        sentence_embeddings = model.encode(sentences)
        # (2)거리 구하는 공식을 적용
        result = np.sqrt(
            np.sum((sentence_embeddings[0]-sentence_embeddings[1])**2))

        if result == 0:
            result = 1
        else:
            result = 1 / result
        testList[i]["similarity"] = result
    return testList