tokenizer.py

import spacy

#spacy.cli.download("en_core_web_sm")

#from spacy.tokens import Doc

# 加载英文模型
#nlp = spacy.load('en_core_web_sm')

import nltk

#nltk.download('punkt')

#from nltk.tokenize import word_tokenize

import jieba

from sacremoses import MosesTokenizer
from subword_nmt import apply_bpe
import codecs

jieba1 = jieba.Tokenizer()
jieba2 = jieba.Tokenizer()
jieba2.load_userdict('model2_data/dict.zh.txt')

mt_zh = MosesTokenizer(lang='zh')
with codecs.open('model2_data/bpecode.zh', 'r', 'utf-8') as f:
    bpe_zh_f = apply_bpe.BPE(f)  

#英文部分初始化，定义tokenize等等
mt_en = MosesTokenizer(lang='en')
with codecs.open('model2_data/bpecode.en', 'r', 'utf-8') as f:
    bpe_en_f = apply_bpe.BPE(f)

def spacy_tokenize(line):
    # 使用spaCy处理文本
    #doc = nlp(line)
    # 获取单词列表
    #words = [token.text for token in doc]
    # 将单词连接成一个字符串，单词间用一个空格间隔
    #return ' '.join(words)
    return ""


def nltk_tokenize(line):
    # 使用NLTK的word_tokenize进行分词
    #tokens = word_tokenize(line)
    #print(tokens)
    #return tokens
    return []


def jieba_tokenize(line):
    # 使用jieba进行分词
    tokens = list(jieba1.cut(line.strip()))  # strip用于去除可能的空白字符
    #print(tokens)
    return tokens

def tokenize(line, mode):
    if mode == "汉译英" :
        return jieba_tokenize(line)
    else :
        return nltk_tokenize(spacy_tokenize(line))


def jieba_tokenize2(line):
    tokens = list(jieba2.cut(line.strip()))
    return tokens

def mt_bpe_zh(line):
    zh_tok = mt_zh.tokenize(line)
    bpe_zh = bpe_zh_f.segment_tokens(zh_tok)
    print(bpe_zh)
    return bpe_zh

def mt_bpe_en(line):
    en_tok = mt_en.tokenize(line)
    bpe_en = bpe_en_f.segment_tokens(en_tok)
    print(bpe_en)
    return bpe_en

def tokenize2(line, mode):
    if mode == "汉译英" :
        return mt_bpe_zh(' '.join(jieba_tokenize2(line)))
    else :
        return mt_bpe_en(line)