{"id":979946,"date":"2024-12-27T06:50:36","date_gmt":"2024-12-26T22:50:36","guid":{"rendered":""},"modified":"2024-12-27T06:50:38","modified_gmt":"2024-12-26T22:50:38","slug":"python%e5%a6%82%e4%bd%95%e7%bc%96%e5%86%99%e7%bf%bb%e8%af%91%e7%ae%97%e6%b3%95","status":"publish","type":"post","link":"https:\/\/docs.pingcode.com\/ask\/979946.html","title":{"rendered":"python\u5982\u4f55\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5"},"content":{"rendered":"

\"python\u5982\u4f55\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5\"<\/p>\n

Python\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5\u4e3b\u8981\u6d89\u53ca\u5230\u6587\u672c\u7684\u89e3\u6790\u4e0e\u8bed\u8a00\u6a21\u578b\u7684\u6784\u5efa\u3001\u5229\u7528\u73b0\u6709\u7684\u7ffb\u8bd1API\u3001\u6784\u5efa\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\u3002<\/strong>\u5bf9\u4e8e\u521d\u5b66\u8005\uff0c\u6700\u7b80\u5355\u7684\u65b9\u6cd5\u662f\u5229\u7528\u73b0\u6709\u7684\u7ffb\u8bd1API\uff0c\u5982Google Translate API\u3001Microsoft Translator API\u7b49\uff0c\u5b83\u4eec\u63d0\u4f9b\u4e86\u73b0\u6210\u7684\u7ffb\u8bd1\u529f\u80fd\uff0c\u6613\u4e8e\u96c6\u6210\u5230Python\u5e94\u7528\u4e2d\u3002\u5bf9\u4e8e\u6709\u4e00\u5b9a\u7f16\u7a0b\u57fa\u7840\u548c\u673a\u5668\u5b66\u4e60<\/a>\u7ecf\u9a8c\u7684\u5f00\u53d1\u8005\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u6784\u5efa\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\uff0c\u4f8b\u5982\u4f7f\u7528TensorFlow\u6216PyTorch\u6846\u67b6\uff0c\u901a\u8fc7\u8bad\u7ec3\u6570\u636e\u6765\u5b9e\u73b0\u7ffb\u8bd1\u529f\u80fd\u3002\u4ee5\u4e0b\u5c06\u8be6\u7ec6\u4ecb\u7ecd\u5176\u4e2d\u7684\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\u3002<\/p>\n<\/p>\n

\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\u901a\u5e38\u4f7f\u7528\u5e8f\u5217\u5230\u5e8f\u5217\uff08Seq2Seq\uff09\u6a21\u578b\uff0c\u8fd9\u79cd\u6a21\u578b\u53ef\u4ee5\u4ece\u6e90\u8bed\u8a00\u5e8f\u5217\u6620\u5c04\u5230\u76ee\u6807\u8bed\u8a00\u5e8f\u5217\u3002Seq2Seq\u6a21\u578b\u5305\u62ec\u7f16\u7801\u5668\uff08encoder\uff09\u548c\u89e3\u7801\u5668\uff08decoder\uff09\u4e24\u4e2a\u4e3b\u8981\u90e8\u5206\u3002\u7f16\u7801\u5668\u7684\u4efb\u52a1\u662f\u5c06\u8f93\u5165\u7684\u6e90\u8bed\u8a00\u53e5\u5b50\u8f6c\u6362\u4e3a\u4e00\u4e2a\u56fa\u5b9a\u957f\u5ea6\u7684\u4e0a\u4e0b\u6587\u5411\u91cf\uff0c\u800c\u89e3\u7801\u5668\u5219\u6839\u636e\u8fd9\u4e2a\u4e0a\u4e0b\u6587\u5411\u91cf\u751f\u6210\u76ee\u6807\u8bed\u8a00\u53e5\u5b50\u3002\u4e3a\u4e86\u63d0\u9ad8\u7ffb\u8bd1\u7684\u51c6\u786e\u6027\uff0c\u5e38\u5e38\u4f7f\u7528\u6ce8\u610f\u529b\u673a\u5236\uff08Attention Mechanism\uff09\u6765\u589e\u5f3a\u6a21\u578b\u7684\u6027\u80fd\u3002\u6ce8\u610f\u529b\u673a\u5236\u5141\u8bb8\u6a21\u578b\u5728\u7ffb\u8bd1\u6bcf\u4e2a\u8bcd\u65f6\u201c\u5173\u6ce8\u201d\u6e90\u53e5\u5b50\u4e2d\u7684\u4e0d\u540c\u90e8\u5206\uff0c\u800c\u4e0d\u662f\u4ec5\u4ec5\u4f9d\u8d56\u4e8e\u56fa\u5b9a\u957f\u5ea6\u7684\u4e0a\u4e0b\u6587\u5411\u91cf\u3002<\/p>\n<\/p>\n

\u4e00\u3001\u5229\u7528\u73b0\u6709\u7ffb\u8bd1API<\/h3>\n<\/p>\n

\u73b0\u6709\u7ffb\u8bd1API\u5982Google Translate API\u548cMicrosoft Translator API\u63d0\u4f9b\u4e86\u5f3a\u5927\u7684\u7ffb\u8bd1\u529f\u80fd\uff0c\u4e14\u6613\u4e8e\u4f7f\u7528\u3002\u5229\u7528\u8fd9\u4e9bAPI\u8fdb\u884c\u7ffb\u8bd1\u662f\u521d\u5b66\u8005\u5feb\u901f\u5b9e\u73b0\u7ffb\u8bd1\u529f\u80fd\u7684\u6709\u6548\u9014\u5f84\u3002<\/p>\n<\/p>\n

1. Google Translate API<\/h4>\n<\/p>\n

Google Translate API\u662f\u4e00\u4e2a\u975e\u5e38\u6d41\u884c\u7684\u7ffb\u8bd1\u670d\u52a1\uff0c\u652f\u6301\u591a\u79cd\u8bed\u8a00\u3002\u4f7f\u7528\u65f6\uff0c\u9700\u8981\u5148\u5728Google Cloud Platform\u4e0a\u542f\u7528Google Translate API\u5e76\u83b7\u53d6API\u5bc6\u94a5\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\uff1a<\/p>\n<\/p>\n

from google.cloud import translate_v2 as translate<\/p>\n
def translate_text(text, target_language):<\/p>\n
    translate_client = translate.Client()<\/p>\n
    result = translate_client.translate(text, target_language=target_language)<\/p>\n
    return result['translatedText']<\/p>\n
translated_text = translate_text("Hello, world!", "es")<\/p>\n
print(translated_text)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u793a\u4f8b\u4e2d\uff0c\u6211\u4eec\u9996\u5148\u521b\u5efa\u4e86\u4e00\u4e2atranslate.Client<\/code>\u5bf9\u8c61\uff0c\u7136\u540e\u8c03\u7528translate<\/code>\u65b9\u6cd5\u8fdb\u884c\u7ffb\u8bd1\u3002text<\/code>\u53c2\u6570\u662f\u8981\u7ffb\u8bd1\u7684\u6587\u672c\uff0c\u800ctarget_language<\/code>\u53c2\u6570\u6307\u5b9a\u76ee\u6807\u8bed\u8a00\u3002<\/p>\n<\/p>\n
2. Microsoft Translator API<\/h4>\n<\/p>\n
Microsoft Translator API\u662f\u53e6\u4e00\u4e2a\u5f3a\u5927\u7684\u7ffb\u8bd1\u670d\u52a1\u3002\u8981\u4f7f\u7528\u6b64API\uff0c\u9996\u5148\u9700\u8981\u5728Microsoft Azure\u95e8\u6237\u4e2d\u521b\u5efa\u4e00\u4e2a\u7ffb\u8bd1\u8d44\u6e90\u5e76\u83b7\u53d6\u5bc6\u94a5\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\uff1a<\/p>\n<\/p>\n
import requests<\/p>\n
def translate_text(text, target_language):<\/p>\n
    subscription_key = 'YOUR_SUBSCRIPTION_KEY'<\/p>\n
    endpoint = 'https:\/\/api.cognitive.microsofttranslator.com'<\/p>\n
    path = '\/translate?api-version=3.0'<\/p>\n
    params = '&to=' + target_language<\/p>\n
    constructed_url = endpoint + path + params<\/p>\n
    headers = {<\/p>\n
        'Ocp-Apim-Subscription-Key': subscription_key,<\/p>\n
        'Content-type': 'application\/json'<\/p>\n
    }<\/p>\n
    body = [{<\/p>\n
        'text': text<\/p>\n
    }]<\/p>\n
    request = requests.post(constructed_url, headers=headers, json=body)<\/p>\n
    response = request.json()<\/p>\n
    return response[0]['translations'][0]['text']<\/p>\n
translated_text = translate_text("Hello, world!", "es")<\/p>\n
print(translated_text)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u6b64\u793a\u4f8b\u4f7f\u7528\u4e86Python\u7684requests<\/code>\u5e93\u5411Microsoft Translator API\u53d1\u9001\u8bf7\u6c42\uff0c\u5e76\u8fd4\u56de\u7ffb\u8bd1\u7ed3\u679c\u3002<\/p>\n<\/p>\n
\u4e8c\u3001\u6784\u5efa\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b<\/h3>\n<\/p>\n
\u5bf9\u4e8e\u5e0c\u671b\u6df1\u5165\u7814\u7a76\u7ffb\u8bd1\u7b97\u6cd5\u7684\u5f00\u53d1\u8005\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u4f7f\u7528\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u5982TensorFlow\u6216PyTorch\u6765\u6784\u5efa\u81ea\u5df1\u7684\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\u3002<\/p>\n<\/p>\n
1. \u5e8f\u5217\u5230\u5e8f\u5217\uff08Seq2Seq\uff09\u6a21\u578b<\/h4>\n<\/p>\n
Seq2Seq\u6a21\u578b\u662f\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u7684\u57fa\u7840\u3002\u5b83\u7531\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u7ec4\u6210\uff0c\u7f16\u7801\u5668\u5c06\u8f93\u5165\u5e8f\u5217\u8f6c\u6362\u4e3a\u4e0a\u4e0b\u6587\u5411\u91cf\uff0c\u89e3\u7801\u5668\u6839\u636e\u4e0a\u4e0b\u6587\u5411\u91cf\u751f\u6210\u8f93\u51fa\u5e8f\u5217\u3002<\/p>\n<\/p>\n
\u7f16\u7801\u5668<\/strong><\/p>\n<\/p>\n
\u7f16\u7801\u5668\u901a\u5e38\u7531\u4e00\u7ec4\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff08RNN\uff09\u3001\u957f\u77ed\u65f6\u8bb0\u5fc6\u7f51\u7edc\uff08LSTM\uff09\u6216\u95e8\u63a7\u5faa\u73af\u5355\u5143\uff08GRU\uff09\u7ec4\u6210\u3002\u7f16\u7801\u5668\u7684\u4efb\u52a1\u662f\u8bfb\u53d6\u8f93\u5165\u53e5\u5b50\u5e76\u5c06\u5176\u8f6c\u6362\u4e3a\u4e0a\u4e0b\u6587\u5411\u91cf\uff0c\u8fd9\u4e2a\u5411\u91cf\u5305\u542b\u4e86\u8f93\u5165\u53e5\u5b50\u7684\u8bed\u4e49\u4fe1\u606f\u3002<\/p>\n<\/p>\n
\u89e3\u7801\u5668<\/strong><\/p>\n<\/p>\n
\u89e3\u7801\u5668\u4e5f\u662f\u7531RNN\u3001LSTM\u6216GRU\u7ec4\u6210\u3002\u89e3\u7801\u5668\u6839\u636e\u7f16\u7801\u5668\u751f\u6210\u7684\u4e0a\u4e0b\u6587\u5411\u91cf\u751f\u6210\u76ee\u6807\u8bed\u8a00\u7684\u53e5\u5b50\u3002\u6bcf\u4e2a\u65f6\u95f4\u6b65\uff0c\u89e3\u7801\u5668\u90fd\u4f1a\u9884\u6d4b\u4e0b\u4e00\u4e2a\u8bcd\uff0c\u76f4\u5230\u751f\u6210\u7ed3\u675f\u6807\u8bb0\u3002<\/p>\n<\/p>\n
\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684Seq2Seq\u6a21\u578b\u7684\u4ee3\u7801\u793a\u4f8b\uff0c\u4f7f\u7528TensorFlow\u6784\u5efa\uff1a<\/p>\n<\/p>\n
import tensorflow as tf<\/p>\n
from tensorflow.keras.layers import Embedding, LSTM, Dense<\/p>\n
from tensorflow.keras.models import Model<\/p>\n
class Encoder(Model):<\/p>\n
    def __init__(self, vocab_size, embedding_dim, enc_units, batch_sz):<\/p>\n
        super(Encoder, self).__init__()<\/p>\n
        self.batch_sz = batch_sz<\/p>\n
        self.enc_units = enc_units<\/p>\n
        self.embedding = Embedding(vocab_size, embedding_dim)<\/p>\n
        self.lstm = LSTM(self.enc_units, return_sequences=True, return_state=True)<\/p>\n
    def call(self, x, hidden):<\/p>\n
        x = self.embedding(x)<\/p>\n
        output, state_h, state_c = self.lstm(x, initial_state=hidden)<\/p>\n
        return output, state_h, state_c<\/p>\n
    def initialize_hidden_state(self):<\/p>\n
        return [tf.zeros((self.batch_sz, self.enc_units)), tf.zeros((self.batch_sz, self.enc_units))]<\/p>\n
class Decoder(Model):<\/p>\n
    def __init__(self, vocab_size, embedding_dim, dec_units, batch_sz):<\/p>\n
        super(Decoder, self).__init__()<\/p>\n
        self.batch_sz = batch_sz<\/p>\n
        self.dec_units = dec_units<\/p>\n
        self.embedding = Embedding(vocab_size, embedding_dim)<\/p>\n
        self.lstm = LSTM(self.dec_units, return_sequences=True, return_state=True)<\/p>\n
        self.fc = Dense(vocab_size)<\/p>\n
    def call(self, x, hidden, enc_output):<\/p>\n
        x = self.embedding(x)<\/p>\n
        output, state_h, state_c = self.lstm(x, initial_state=hidden)<\/p>\n
        output = self.fc(output)<\/p>\n
        return output, state_h, state_c<\/p>\n
Instantiate encoder and decoder<\/strong><\/h2>\n
encoder = Encoder(vocab_size=10000, embedding_dim=256, enc_units=512, batch_sz=64)<\/p>\n
decoder = Decoder(vocab_size=10000, embedding_dim=256, dec_units=512, batch_sz=64)<\/p>\n
<\/code><\/pre>\n<\/p>\n
2. \u6ce8\u610f\u529b\u673a\u5236<\/h4>\n<\/p>\n
\u6ce8\u610f\u529b\u673a\u5236\u53ef\u4ee5\u663e\u8457\u63d0\u5347Seq2Seq\u6a21\u578b\u7684\u6027\u80fd\u3002\u5b83\u5141\u8bb8\u89e3\u7801\u5668\u5728\u751f\u6210\u6bcf\u4e2a\u8bcd\u65f6\uff0c\u9009\u62e9\u6027\u5730\u5173\u6ce8\u7f16\u7801\u5668\u8f93\u51fa\u4e2d\u7684\u4e0d\u540c\u90e8\u5206\uff0c\u800c\u4e0d\u662f\u7b80\u5355\u5730\u4f9d\u8d56\u56fa\u5b9a\u7684\u4e0a\u4e0b\u6587\u5411\u91cf\u3002<\/p>\n<\/p>\n
\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u6ce8\u610f\u529b\u673a\u5236\u7684\u5b9e\u73b0\u793a\u4f8b\uff1a<\/p>\n<\/p>\n
class BahdanauAttention(tf.keras.layers.Layer):<\/p>\n
    def __init__(self, units):<\/p>\n
        super(BahdanauAttention, self).__init__()<\/p>\n
        self.W1 = Dense(units)<\/p>\n
        self.W2 = Dense(units)<\/p>\n
        self.V = Dense(1)<\/p>\n
    def call(self, query, values):<\/p>\n
        hidden_with_time_axis = tf.expand_dims(query, 1)<\/p>\n
        score = self.V(tf.nn.tanh(self.W1(values) + self.W2(hidden_with_time_axis)))<\/p>\n
        attention_weights = tf.nn.softmax(score, axis=1)<\/p>\n
        context_vector = attention_weights * values<\/p>\n
        context_vector = tf.reduce_sum(context_vector, axis=1)<\/p>\n
        return context_vector, attention_weights<\/p>\n
attention_layer = BahdanauAttention(units=512)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u4f7f\u7528\u6ce8\u610f\u529b\u673a\u5236\u65f6\uff0c\u89e3\u7801\u5668\u6bcf\u4e2a\u65f6\u95f4\u6b65\u4f1a\u5229\u7528\u6ce8\u610f\u529b\u5c42\u8ba1\u7b97\u5f53\u524d\u65f6\u95f4\u6b65\u7684\u4e0a\u4e0b\u6587\u5411\u91cf\uff0c\u8fd9\u4e00\u5411\u91cf\u7ed3\u5408\u4e86\u7f16\u7801\u5668\u8f93\u51fa\u7684\u4e0d\u540c\u90e8\u5206\u7684\u4fe1\u606f\u3002<\/p>\n<\/p>\n
\u4e09\u3001\u6570\u636e\u51c6\u5907\u4e0e\u8bad\u7ec3<\/h3>\n<\/p>\n
\u8981\u8bad\u7ec3\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\uff0c\u9996\u5148\u9700\u8981\u51c6\u5907\u597d\u8bad\u7ec3\u6570\u636e\u3002\u901a\u5e38\u4f7f\u7528\u53cc\u8bed\u5e73\u884c\u8bed\u6599\u5e93\uff08parallel corpus\uff09\uff0c\u5b83\u5305\u542b\u6210\u5bf9\u7684\u6e90\u8bed\u8a00\u548c\u76ee\u6807\u8bed\u8a00\u53e5\u5b50\u3002\u5e38\u7528\u7684\u6570\u636e\u96c6\u5305\u62ecWMT\u3001IWSLT\u7b49\u3002<\/p>\n<\/p>\n
1. \u6570\u636e\u9884\u5904\u7406<\/h4>\n<\/p>\n
\u5728\u8bad\u7ec3\u6a21\u578b\u4e4b\u524d\uff0c\u5fc5\u987b\u5bf9\u6570\u636e\u8fdb\u884c\u9884\u5904\u7406\uff0c\u5305\u62ec\u6807\u8bb0\u5316\u3001\u53bb\u9664\u6807\u70b9\u7b26\u53f7\u3001\u8f6c\u6362\u4e3a\u5c0f\u5199\u7b49\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\uff1a<\/p>\n<\/p>\n
import re<\/p>\n
def preprocess_sentence(sentence):<\/p>\n
    sentence = sentence.lower().strip()<\/p>\n
    sentence = re.sub(r"([?.!,\u00bf])", r" \\1 ", sentence)<\/p>\n
    sentence = re.sub(r'[" "]+', " ", sentence)<\/p>\n
    sentence = re.sub(r"[^a-zA-Z?.!,\u00bf]+", " ", sentence)<\/p>\n
    sentence = sentence.rstrip().strip()<\/p>\n
    sentence = '<start> ' + sentence + ' <end>'<\/p>\n
    return sentence<\/p>\n
<\/code><\/pre>\n<\/p>\n
2. \u6a21\u578b\u8bad\u7ec3<\/h4>\n<\/p>\n
\u5728\u6570\u636e\u9884\u5904\u7406\u5b8c\u6210\u540e\uff0c\u53ef\u4ee5\u5f00\u59cb\u8bad\u7ec3\u6a21\u578b\u3002\u8bad\u7ec3\u8fc7\u7a0b\u901a\u5e38\u5305\u62ec\u5b9a\u4e49\u635f\u5931\u51fd\u6570\u3001\u4f18\u5316\u5668\u548c\u8bad\u7ec3\u5faa\u73af\u3002<\/p>\n<\/p>\n
optimizer = tf.keras.optimizers.Adam()<\/p>\n
loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True, reduction='none')<\/p>\n
def loss_function(real, pred):<\/p>\n
    mask = tf.math.logical_not(tf.math.equal(real, 0))<\/p>\n
    loss_ = loss_object(real, pred)<\/p>\n
    mask = tf.cast(mask, dtype=loss_.dtype)<\/p>\n
    loss_ *= mask<\/p>\n
    return tf.reduce_mean(loss_)<\/p>\n
@tf.function<\/p>\n
def trAI<\/a>n_step(inp, targ, enc_hidden):<\/p>\n
    loss = 0<\/p>\n
    with tf.GradientTape() as tape:<\/p>\n
        enc_output, enc_hidden_h, enc_hidden_c = encoder(inp, enc_hidden)<\/p>\n
        dec_hidden = [enc_hidden_h, enc_hidden_c]<\/p>\n
        dec_input = tf.expand_dims([targ_lang.word_index['<start>']] * BATCH_SIZE, 1)<\/p>\n
        for t in range(1, targ.shape[1]):<\/p>\n
            predictions, dec_hidden_h, dec_hidden_c = decoder(dec_input, dec_hidden, enc_output)<\/p>\n
            loss += loss_function(targ[:, t], predictions)<\/p>\n
            dec_input = tf.expand_dims(targ[:, t], 1)<\/p>\n
    batch_loss = (loss \/ int(targ.shape[1]))<\/p>\n
    variables = encoder.trainable_variables + decoder.trainable_variables<\/p>\n
    gradients = tape.gradient(loss, variables)<\/p>\n
    optimizer.apply_gradients(zip(gradients, variables))<\/p>\n
    return batch_loss<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u56db\u3001\u6a21\u578b\u8bc4\u4f30\u4e0e\u4f18\u5316<\/h3>\n<\/p>\n
\u5728\u6a21\u578b\u8bad\u7ec3\u5b8c\u6210\u540e\uff0c\u9700\u8981\u5bf9\u6a21\u578b\u8fdb\u884c\u8bc4\u4f30\u548c\u4f18\u5316\uff0c\u4ee5\u786e\u4fdd\u5176\u5728\u7ffb\u8bd1\u4efb\u52a1\u4e0a\u7684\u6027\u80fd\u3002<\/p>\n<\/p>\n
1. \u6a21\u578b\u8bc4\u4f30<\/h4>\n<\/p>\n
\u6a21\u578b\u8bc4\u4f30\u901a\u5e38\u5305\u62ec\u8ba1\u7b97BLEU\u5206\u6570\u3001\u51c6\u786e\u7387\u7b49\u6307\u6807\u3002BLEU\u5206\u6570\u662f\u4e00\u79cd\u5e38\u7528\u7684\u673a\u5668\u7ffb\u8bd1\u8d28\u91cf\u8bc4\u4f30\u6307\u6807\uff0c\u5b83\u901a\u8fc7\u6bd4\u8f83\u6a21\u578b\u751f\u6210\u7684\u7ffb\u8bd1\u4e0e\u53c2\u8003\u7ffb\u8bd1\u4e4b\u95f4\u7684n\u5143\u7ec4\u91cd\u5408\u7a0b\u5ea6\u6765\u8ba1\u7b97\u5f97\u5206\u3002<\/p>\n<\/p>\n
2. \u6a21\u578b\u4f18\u5316<\/h4>\n<\/p>\n
\u4e3a\u4e86\u8fdb\u4e00\u6b65\u63d0\u9ad8\u6a21\u578b\u7684\u6027\u80fd\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u8fdb\u884c\u4ee5\u4e0b\u4f18\u5316\uff1a<\/p>\n<\/p>\n
    \n
  • \u589e\u52a0\u6a21\u578b\u590d\u6742\u5ea6<\/strong>\uff1a\u589e\u52a0\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u7684\u5c42\u6570\u3001\u4f7f\u7528\u66f4\u5927\u7684\u9690\u85cf\u5355\u5143\u7b49\u3002<\/li>\n
  • \u6570\u636e\u589e\u5f3a<\/strong>\uff1a\u4f7f\u7528\u6570\u636e\u589e\u5f3a\u6280\u672f\uff0c\u5982\u53e5\u5b50\u91cd\u6392\u3001\u540c\u4e49\u8bcd\u66ff\u6362\u7b49\uff0c\u589e\u52a0\u8bad\u7ec3\u6570\u636e\u7684\u591a\u6837\u6027\u3002<\/li>\n
  • \u8d85\u53c2\u6570\u4f18\u5316<\/strong>\uff1a\u8c03\u6574\u5b66\u4e60\u7387\u3001\u6279\u5927\u5c0f\u7b49\u8d85\u53c2\u6570\uff0c\u5bfb\u627e\u6700\u4f73\u7684\u8bad\u7ec3\u914d\u7f6e\u3002<\/li>\n<\/ul>\n
    \u901a\u8fc7\u4e0a\u8ff0\u65b9\u6cd5\uff0c\u53ef\u4ee5\u6709\u6548\u5730\u5229\u7528Python\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5\uff0c\u5e76\u5b9e\u73b0\u9ad8\u6548\u7684\u673a\u5668\u7ffb\u8bd1\u7cfb\u7edf\u3002\u65e0\u8bba\u662f\u4f7f\u7528\u73b0\u6709\u7684API\u8fd8\u662f\u6784\u5efa\u81ea\u5b9a\u4e49\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u5173\u952e\u5728\u4e8e\u7406\u89e3\u7ffb\u8bd1\u8fc7\u7a0b\u4e2d\u7684\u6838\u5fc3\u6280\u672f\u548c\u65b9\u6cd5\uff0c\u4ece\u800c\u6839\u636e\u5177\u4f53\u9700\u6c42\u9009\u62e9\u5408\u9002\u7684\u65b9\u6848\u3002<\/p>\n<\/p>\n
    \u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n
     \u5982\u4f55\u9009\u62e9\u5408\u9002\u7684\u7ffb\u8bd1\u5e93\u6216\u5de5\u5177\uff1f<\/strong>
    \u5728\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5\u65f6\uff0c\u9009\u62e9\u9002\u5408\u7684\u5e93\u6216\u5de5\u5177\u81f3\u5173\u91cd\u8981\u3002Python\u4e2d\u6709\u591a\u79cd\u7ffb\u8bd1\u5e93\u53ef\u4f9b\u4f7f\u7528\uff0c\u4f8b\u5982Google Translate API\u3001DeepL API\u548cMicrosoft Translator\u3002\u4e0d\u540c\u7684\u5e93\u5728\u8bed\u8a00\u652f\u6301\u3001\u7ffb\u8bd1\u8d28\u91cf\u548c\u4f7f\u7528\u6210\u672c\u4e0a\u5404\u6709\u7279\u70b9\u3002\u53ef\u4ee5\u6839\u636e\u9879\u76ee\u7684\u9700\u6c42\u3001\u9884\u7b97\u548c\u6280\u672f\u6808\u6765\u9009\u62e9\u6700\u5408\u9002\u7684\u5de5\u5177\u3002<\/p>\n
    \u5982\u4f55\u5904\u7406\u7ffb\u8bd1\u4e2d\u7684\u8bed\u5883\u548c\u6b67\u4e49\u95ee\u9898\uff1f<\/strong>
    \u7ffb\u8bd1\u7b97\u6cd5\u5728\u5904\u7406\u8bed\u5883\u548c\u6b67\u4e49\u65f6\u9762\u4e34\u6311\u6218\u3002\u4e3a\u4e86\u63d0\u5347\u7ffb\u8bd1\u8d28\u91cf\uff0c\u53ef\u4ee5\u5f15\u5165\u4e0a\u4e0b\u6587\u5206\u6790\u3002\u901a\u8fc7\u4f7f\u7528\u81ea\u7136\u8bed\u8a00\u5904\u7406\u6280\u672f\uff08\u5982\u8bcd\u6027\u6807\u6ce8\u548c\u547d\u540d\u5b9e\u4f53\u8bc6\u522b\uff09\uff0c\u53ef\u4ee5\u66f4\u597d\u5730\u7406\u89e3\u53e5\u5b50\u7684\u542b\u4e49\uff0c\u4ece\u800c\u51cf\u5c11\u7ffb\u8bd1\u4e2d\u7684\u8bef\u89e3\u3002\u6b64\u5916\uff0c\u5229\u7528\u673a\u5668\u5b66\u4e60\u6a21\u578b\uff0c\u5c24\u5176\u662f\u8bad\u7ec3\u826f\u597d\u7684\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\uff0c\u53ef\u4ee5\u663e\u8457\u63d0\u9ad8\u5bf9\u590d\u6742\u53e5\u5b50\u7684\u7ffb\u8bd1\u51c6\u786e\u6027\u3002<\/p>\n
    \u5982\u4f55\u8bc4\u4f30\u7ffb\u8bd1\u7b97\u6cd5\u7684\u51c6\u786e\u6027\u548c\u6027\u80fd\uff1f<\/strong>
    \u8bc4\u4f30\u7ffb\u8bd1\u7b97\u6cd5\u7684\u6548\u679c\u53ef\u4ee5\u91c7\u7528\u591a\u79cd\u65b9\u6cd5\uff0c\u5305\u62ec\u4eba\u5de5\u8bc4\u5ba1\u548c\u81ea\u52a8\u5316\u8bc4\u4f30\u6307\u6807\u3002\u5e38\u7528\u7684\u81ea\u52a8\u5316\u8bc4\u4f30\u6307\u6807\u6709BLEU\u3001ROUGE\u548cMETEOR\u7b49\uff0c\u5b83\u4eec\u901a\u8fc7\u6bd4\u8f83\u751f\u6210\u7684\u7ffb\u8bd1\u4e0e\u53c2\u8003\u7ffb\u8bd1\u4e4b\u95f4\u7684\u76f8\u4f3c\u5ea6\u6765\u91cf\u5316\u7ffb\u8bd1\u8d28\u91cf\u3002\u540c\u65f6\uff0c\u8fdb\u884c\u7528\u6237\u6d4b\u8bd5\u4e5f\u662f\u4e00\u4e2a\u6709\u6548\u7684\u65b9\u6cd5\uff0c\u6536\u96c6\u7528\u6237\u53cd\u9988\u53ef\u4ee5\u5e2e\u52a9\u8bc6\u522b\u6f5c\u5728\u7684\u95ee\u9898\u548c\u6539\u8fdb\u65b9\u5411\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"Python\u7f16\u5199\u7ffb\u8bd1\u7b97\u6cd5\u4e3b\u8981\u6d89\u53ca\u5230\u6587\u672c\u7684\u89e3\u6790\u4e0e\u8bed\u8a00\u6a21\u578b\u7684\u6784\u5efa\u3001\u5229\u7528\u73b0\u6709\u7684\u7ffb\u8bd1API\u3001\u6784\u5efa\u795e\u7ecf\u7f51\u7edc\u7ffb\u8bd1\u6a21\u578b\u3002\u5bf9\u4e8e […]","protected":false},"author":3,"featured_media":979957,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[37],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/979946"}],"collection":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/comments?post=979946"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/979946\/revisions"}],"predecessor-version":[{"id":979960,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/979946\/revisions\/979960"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/979957"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=979946"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=979946"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=979946"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}