{"id":1027744,"date":"2024-12-31T10:53:35","date_gmt":"2024-12-31T02:53:35","guid":{"rendered":"https:\/\/docs.pingcode.com\/ask\/ask-ask\/1027744.html"},"modified":"2024-12-31T10:53:37","modified_gmt":"2024-12-31T02:53:37","slug":"python%e5%a6%82%e4%bd%95%e6%8f%90%e5%8f%96%e8%af%84%e8%ae%ba%e4%b8%ad%e5%85%b3%e9%94%ae%e8%af%8d","status":"publish","type":"post","link":"https:\/\/docs.pingcode.com\/ask\/1027744.html","title":{"rendered":"python\u5982\u4f55\u63d0\u53d6\u8bc4\u8bba\u4e2d\u5173\u952e\u8bcd"},"content":{"rendered":"

\"python\u5982\u4f55\u63d0\u53d6\u8bc4\u8bba\u4e2d\u5173\u952e\u8bcd\"<\/p>\n

Python \u63d0\u53d6\u8bc4\u8bba\u4e2d\u5173\u952e\u8bcd\u7684\u65b9\u6cd5\u5305\u62ec\u4f7f\u7528\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u5de5\u5177\u548c\u5e93\uff0c\u5982NLTK\u3001spaCy\u4ee5\u53caTF-IDF\u7b49\u3002\u9996\u5148\uff0c\u901a\u5e38\u9700\u8981\u5bf9\u8bc4\u8bba\u8fdb\u884c\u9884\u5904\u7406\uff0c\u5305\u62ec\u53bb\u9664\u505c\u7528\u8bcd\u3001\u8bcd\u5e72\u63d0\u53d6\u7b49\u3002\u63a5\u7740\uff0c\u53ef\u4ee5\u5229\u7528TF-IDF\u7b97\u6cd5\u3001\u4e3b\u9898\u6a21\u578b\u6216\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\u7b49\u65b9\u5f0f\u6765\u63d0\u53d6\u5173\u952e\u8bcd\u3002<\/strong><\/p>\n<\/p>\n

\u5176\u4e2d\u4e00\u4e2a\u5e38\u7528\u7684\u65b9\u6cd5\u662fTF-IDF\u7b97\u6cd5<\/strong>\uff0c\u5b83\u80fd\u8861\u91cf\u4e00\u4e2a\u8bcd\u5728\u6587\u6863\u4e2d\u7684\u91cd\u8981\u6027\uff0c\u5e76\u901a\u8fc7\u8ba1\u7b97\u8bcd\u9891\u548c\u9006\u6587\u6863\u9891\u7387\u6765\u786e\u5b9a\u5173\u952e\u8bcd\u3002\u672c\u6587\u5c06\u8be6\u7ec6\u4ecb\u7ecd\u5982\u4f55\u4f7f\u7528Python\u53ca\u76f8\u5173\u5e93\u6765\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u3002<\/p>\n<\/p>\n

\u4e00\u3001\u6587\u672c\u9884\u5904\u7406<\/h3>\n<\/p>\n

\u5728\u8fdb\u884c\u5173\u952e\u8bcd\u63d0\u53d6\u4e4b\u524d\uff0c\u9996\u5148\u9700\u8981\u5bf9\u8bc4\u8bba\u6587\u672c\u8fdb\u884c\u9884\u5904\u7406\u3002\u9884\u5904\u7406\u7684\u6b65\u9aa4\u5305\u62ec\u6587\u672c\u6e05\u6d17\u3001\u53bb\u9664\u505c\u7528\u8bcd\u3001\u8bcd\u5e72\u63d0\u53d6\u7b49\u3002<\/p>\n<\/p>\n

1\u3001\u6587\u672c\u6e05\u6d17<\/h4>\n<\/p>\n

\u6587\u672c\u6e05\u6d17\u662f\u5bf9\u539f\u59cb\u6587\u672c\u8fdb\u884c\u89c4\u8303\u5316\u5904\u7406\u7684\u8fc7\u7a0b\uff0c\u901a\u5e38\u5305\u62ec\u53bb\u9664\u6807\u70b9\u7b26\u53f7\u3001\u6570\u5b57\u3001\u7279\u6b8a\u5b57\u7b26\u7b49\u3002<\/p>\n<\/p>\n

import re<\/p>\n
def clean_text(text):<\/p>\n
    # \u53bb\u9664\u6807\u70b9\u7b26\u53f7\u3001\u6570\u5b57\u548c\u7279\u6b8a\u5b57\u7b26<\/p>\n
    text = re.sub(r'[^\\w\\s]', '', text)<\/p>\n
    text = re.sub(r'\\d+', '', text)<\/p>\n
    text = text.lower()<\/p>\n
    return text<\/p>\n
<\/code><\/pre>\n<\/p>\n
2\u3001\u53bb\u9664\u505c\u7528\u8bcd<\/h4>\n<\/p>\n
\u505c\u7528\u8bcd\u662f\u6307\u5728\u6587\u672c\u4e2d\u51fa\u73b0\u9891\u7387\u8f83\u9ad8\u4f46\u5bf9\u6587\u672c\u5185\u5bb9\u8d21\u732e\u4e0d\u5927\u7684\u8bcd\u8bed\uff0c\u5982\u201c\u7684\u201d\u3001\u201c\u662f\u201d\u3001\u201c\u5728\u201d\u7b49\u3002\u53ef\u4ee5\u4f7f\u7528NLTK\u5e93\u4e2d\u7684\u505c\u7528\u8bcd\u5217\u8868\u8fdb\u884c\u53bb\u9664\u3002<\/p>\n<\/p>\n
from nltk.corpus import stopwords<\/p>\n
stop_words = set(stopwords.words('english'))<\/p>\n
def remove_stopwords(text):<\/p>\n
    return ' '.join([word for word in text.split() if word not in stop_words])<\/p>\n
<\/code><\/pre>\n<\/p>\n
3\u3001\u8bcd\u5e72\u63d0\u53d6<\/h4>\n<\/p>\n
\u8bcd\u5e72\u63d0\u53d6\u662f\u5c06\u8bcd\u8bed\u8fd8\u539f\u4e3a\u5176\u8bcd\u6839\u5f62\u5f0f\u7684\u8fc7\u7a0b\uff0c\u53ef\u4ee5\u4f7f\u7528NLTK\u5e93\u4e2d\u7684PorterStemmer\u6216LancasterStemmer\u6765\u5b9e\u73b0\u3002<\/p>\n<\/p>\n
from nltk.stem import PorterStemmer<\/p>\n
stemmer = PorterStemmer()<\/p>\n
def stem_words(text):<\/p>\n
    return ' '.join([stemmer.stem(word) for word in text.split()])<\/p>\n
<\/code><\/pre>\n<\/p>\n
4\u3001\u7efc\u5408\u9884\u5904\u7406<\/h4>\n<\/p>\n
\u5c06\u4ee5\u4e0a\u6b65\u9aa4\u7efc\u5408\u8d77\u6765\uff0c\u5bf9\u8bc4\u8bba\u8fdb\u884c\u5168\u9762\u7684\u9884\u5904\u7406\u3002<\/p>\n<\/p>\n
def preprocess_text(text):<\/p>\n
    text = clean_text(text)<\/p>\n
    text = remove_stopwords(text)<\/p>\n
    text = stem_words(text)<\/p>\n
    return text<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e8c\u3001TF-IDF\u7b97\u6cd5\u63d0\u53d6\u5173\u952e\u8bcd<\/h3>\n<\/p>\n
TF-IDF\uff08Term Frequency-Inverse Document Frequency\uff09\u662f\u8861\u91cf\u4e00\u4e2a\u8bcd\u5728\u6587\u6863\u4e2d\u7684\u91cd\u8981\u6027\u7684\u65b9\u6cd5\u3002\u4f7f\u7528sklearn\u5e93\u53ef\u4ee5\u8f7b\u677e\u5b9e\u73b0TF-IDF\u7b97\u6cd5\u3002<\/p>\n<\/p>\n
1\u3001\u8ba1\u7b97TF-IDF\u503c<\/h4>\n<\/p>\n
from sklearn.feature_extraction.text import TfidfVectorizer<\/p>\n
def compute_tfidf(corpus):<\/p>\n
    vectorizer = TfidfVectorizer()<\/p>\n
    tfidf_matrix = vectorizer.fit_transform(corpus)<\/p>\n
    feature_names = vectorizer.get_feature_names_out()<\/p>\n
    return tfidf_matrix, feature_names<\/p>\n
<\/code><\/pre>\n<\/p>\n
2\u3001\u63d0\u53d6\u5173\u952e\u8bcd<\/h4>\n<\/p>\n
\u6839\u636eTF-IDF\u503c\u63d0\u53d6\u6bcf\u6761\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u3002<\/p>\n<\/p>\n
import numpy as np<\/p>\n
def extract_keywords(tfidf_matrix, feature_names, top_n=5):<\/p>\n
    keywords = []<\/p>\n
    for row in tfidf_matrix:<\/p>\n
        sorted_indices = np.argsort(row.toarray()).flatten()[::-1]<\/p>\n
        top_keywords = [feature_names[i] for i in sorted_indices[:top_n]]<\/p>\n
        keywords.append(top_keywords)<\/p>\n
    return keywords<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e09\u3001\u4e3b\u9898\u6a21\u578b\u63d0\u53d6\u5173\u952e\u8bcd<\/h3>\n<\/p>\n
\u9664\u4e86TF-IDF\u7b97\u6cd5\uff0c\u8fd8\u53ef\u4ee5\u4f7f\u7528\u4e3b\u9898\u6a21\u578b\uff08\u5982LDA\uff09\u6765\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u3002<\/p>\n<\/p>\n
1\u3001\u4f7f\u7528Gensim\u5e93\u5b9e\u73b0LDA\u6a21\u578b<\/h4>\n<\/p>\n
import gensim<\/p>\n
from gensim import corpora<\/p>\n
def build_lda_model(corpus, num_topics=5):<\/p>\n
    dictionary = corpora.Dictionary(corpus)<\/p>\n
    doc_term_matrix = [dictionary.doc2bow(doc) for doc in corpus]<\/p>\n
    lda_model = gensim.models.ldamodel.LdaModel(doc_term_matrix, num_topics=num_topics, id2word=dictionary, passes=15)<\/p>\n
    return lda_model, dictionary<\/p>\n
<\/code><\/pre>\n<\/p>\n
2\u3001\u63d0\u53d6\u5173\u952e\u8bcd<\/h4>\n<\/p>\n
\u4eceLDA\u6a21\u578b\u4e2d\u63d0\u53d6\u4e3b\u9898\u548c\u5173\u952e\u8bcd\u3002<\/p>\n<\/p>\n
def extract_lda_keywords(lda_model, dictionary, num_words=5):<\/p>\n
    topics = lda_model.print_topics(num_words=num_words)<\/p>\n
    keywords = []<\/p>\n
    for topic in topics:<\/p>\n
        topic_keywords = [dictionary[int(word.split('*')[1])] for word in topic[1].split('+')]<\/p>\n
        keywords.append(topic_keywords)<\/p>\n
    return keywords<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u56db\u3001\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\u63d0\u53d6\u5173\u952e\u8bcd<\/h3>\n<\/p>\n
\u4f7f\u7528\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\uff08\u5982BERT\uff09\u4e5f\u53ef\u4ee5\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u3002<\/p>\n<\/p>\n
1\u3001\u5b89\u88c5\u548c\u5bfc\u5165Transformers\u5e93<\/h4>\n<\/p>\n
pip install transformers<\/p>\n
<\/code><\/pre>\n<\/p>\n
from transformers import BertTokenizer, BertForTokenClassification<\/p>\n
import torch<\/p>\n
tokenizer = BertTokenizer.from_pretrAI<\/a>ned('bert-base-uncased')<\/p>\n
model = BertForTokenClassification.from_pretrained('bert-base-uncased')<\/p>\n
<\/code><\/pre>\n<\/p>\n
2\u3001\u63d0\u53d6\u5173\u952e\u8bcd<\/h4>\n<\/p>\n
def extract_bert_keywords(text, top_n=5):<\/p>\n
    inputs = tokenizer(text, return_tensors="pt")<\/p>\n
    outputs = model(inputs).logits<\/p>\n
    predictions = torch.argmax(outputs, dim=2)<\/p>\n
    tokens = tokenizer.convert_ids_to_tokens(inputs['input_ids'][0])<\/p>\n
    keywords = [tokens[i] for i in range(len(tokens)) if predictions[0][i] == 1]<\/p>\n
    return keywords[:top_n]<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e94\u3001\u7efc\u5408\u63d0\u53d6\u5173\u952e\u8bcd<\/h3>\n<\/p>\n
\u53ef\u4ee5\u5c06\u4ee5\u4e0a\u65b9\u6cd5\u7efc\u5408\u8d77\u6765\uff0c\u5f62\u6210\u4e00\u4e2a\u5168\u9762\u7684\u5173\u952e\u8bcd\u63d0\u53d6\u6d41\u7a0b\u3002<\/p>\n<\/p>\n
def extract_keywords_from_comments(comments, method='tfidf', top_n=5):<\/p>\n
    preprocessed_comments = [preprocess_text(comment) for comment in comments]<\/p>\n
    if method == 'tfidf':<\/p>\n
        tfidf_matrix, feature_names = compute_tfidf(preprocessed_comments)<\/p>\n
        return extract_keywords(tfidf_matrix, feature_names, top_n)<\/p>\n
    elif method == 'lda':<\/p>\n
        corpus = [comment.split() for comment in preprocessed_comments]<\/p>\n
        lda_model, dictionary = build_lda_model(corpus)<\/p>\n
        return extract_lda_keywords(lda_model, dictionary, top_n)<\/p>\n
    elif method == 'bert':<\/p>\n
        return [extract_bert_keywords(comment, top_n) for comment in comments]<\/p>\n
    else:<\/p>\n
        raise ValueError("Unsupported method. Choose 'tfidf', 'lda', or 'bert'.")<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u516d\u3001\u603b\u7ed3<\/h3>\n<\/p>\n
\u901a\u8fc7\u672c\u6587\u4ecb\u7ecd\u7684\u51e0\u79cd\u65b9\u6cd5\uff0c\u53ef\u4ee5\u6709\u6548\u5730\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u3002TF-IDF\u7b97\u6cd5\u9002\u7528\u4e8e\u6587\u672c\u91cf\u8f83\u5c0f\u7684\u60c5\u51b5<\/strong>\uff0cLDA\u6a21\u578b\u9002\u7528\u4e8e\u4e3b\u9898\u5206\u6790<\/strong>\uff0c\u800c\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\uff08\u5982BERT\uff09\u5219\u9002\u7528\u4e8e\u66f4\u590d\u6742\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1<\/strong>\u3002\u6839\u636e\u5177\u4f53\u9700\u6c42\u9009\u62e9\u5408\u9002\u7684\u65b9\u6cd5\uff0c\u53ef\u4ee5\u63d0\u9ad8\u5173\u952e\u8bcd\u63d0\u53d6\u7684\u6548\u679c\u3002<\/p>\n<\/p>\n
\u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n
 \u5982\u4f55\u4f7f\u7528Python\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\uff1f<\/strong>
\u4f7f\u7528Python\u63d0\u53d6\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u901a\u5e38\u53ef\u4ee5\u5229\u7528\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\uff0c\u6bd4\u5982NLTK\u3001spaCy\u6216\u8005jieba\u3002\u9996\u5148\uff0c\u9700\u8981\u5bf9\u8bc4\u8bba\u8fdb\u884c\u9884\u5904\u7406\uff0c\u5305\u62ec\u53bb\u9664\u505c\u7528\u8bcd\u3001\u6807\u70b9\u7b26\u53f7\u548c\u8fdb\u884c\u5206\u8bcd\u3002\u63a5\u7740\uff0c\u53ef\u4ee5\u4f7f\u7528TF-IDF\uff08\u8bcd\u9891-\u9006\u6587\u6863\u9891\u7387\uff09\u6216\u8005TextRank\u7b49\u7b97\u6cd5\u6765\u8bc6\u522b\u8bc4\u8bba\u4e2d\u7684\u91cd\u8981\u5173\u952e\u8bcd\u3002\u8fd9\u4e9b\u65b9\u6cd5\u53ef\u4ee5\u5e2e\u52a9\u4f60\u627e\u5230\u5728\u8bc4\u8bba\u4e2d\u9891\u7e41\u51fa\u73b0\u4e14\u5177\u6709\u4ee3\u8868\u6027\u7684\u8bcd\u6c47\u3002<\/p>\n
\u63d0\u53d6\u5173\u952e\u8bcd\u65f6\uff0c\u6709\u54ea\u4e9b\u5e38\u7528\u7684Python\u5e93\u63a8\u8350\uff1f<\/strong>
\u5728\u63d0\u53d6\u5173\u952e\u8bcd\u65f6\uff0cNLTK\u3001spaCy\u548cjieba\u662f\u975e\u5e38\u6d41\u884c\u7684Python\u5e93\u3002NLTK\u63d0\u4f9b\u4e86\u4e30\u5bcc\u7684\u6587\u672c\u5904\u7406\u5de5\u5177\uff0c\u9002\u5408\u8fdb\u884c\u57fa\u7840\u7684\u6587\u672c\u5206\u6790\u3002spaCy\u5219\u66f4\u6ce8\u91cd\u901f\u5ea6\u548c\u6548\u7387\uff0c\u9002\u5408\u5904\u7406\u5927\u89c4\u6a21\u6587\u672c\u6570\u636e\u3002jieba\u662f\u4e13\u4e3a\u4e2d\u6587\u6587\u672c\u8bbe\u8ba1\u7684\u5206\u8bcd\u5e93\uff0c\u80fd\u591f\u6709\u6548\u5904\u7406\u4e2d\u6587\u8bc4\u8bba\u4e2d\u7684\u5173\u952e\u8bcd\u63d0\u53d6\u9700\u6c42\u3002<\/p>\n
\u63d0\u53d6\u5173\u952e\u8bcd\u7684\u8fc7\u7a0b\u4e2d\u9700\u8981\u6ce8\u610f\u54ea\u4e9b\u95ee\u9898\uff1f<\/strong>
\u5728\u63d0\u53d6\u5173\u952e\u8bcd\u65f6\uff0c\u9700\u8981\u5173\u6ce8\u8bc4\u8bba\u7684\u4e0a\u4e0b\u6587\u548c\u8bed\u5883\u3002\u4e0d\u540c\u7684\u8bc4\u8bba\u53ef\u80fd\u5728\u7279\u5b9a\u60c5\u51b5\u4e0b\u4f1a\u4ea7\u751f\u4e0d\u540c\u7684\u5173\u952e\u8bcd\u3002\u56e0\u6b64\uff0c\u4f7f\u7528\u60c5\u611f\u5206\u6790\u5de5\u5177\u8f85\u52a9\u63d0\u53d6\u5173\u952e\u8bcd\u4e5f\u662f\u4e00\u4e2a\u4e0d\u9519\u7684\u9009\u62e9\u3002\u6b64\u5916\uff0c\u786e\u4fdd\u53bb\u9664\u505c\u7528\u8bcd\u80fd\u591f\u63d0\u9ad8\u5173\u952e\u8bcd\u63d0\u53d6\u7684\u51c6\u786e\u6027\u3002\u6700\u540e\uff0c\u6570\u636e\u7684\u9884\u5904\u7406\u548c\u6e05\u6d17\u4e5f\u662f\u4e0d\u53ef\u5ffd\u89c6\u7684\u6b65\u9aa4\uff0c\u8fd9\u5c06\u76f4\u63a5\u5f71\u54cd\u5173\u952e\u8bcd\u63d0\u53d6\u7684\u6548\u679c\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"Python \u63d0\u53d6\u8bc4\u8bba\u4e2d\u5173\u952e\u8bcd\u7684\u65b9\u6cd5\u5305\u62ec\u4f7f\u7528\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u5de5\u5177\u548c\u5e93\uff0c\u5982NLTK\u3001spaCy\u4ee5\u53caTF-I […]","protected":false},"author":3,"featured_media":1027760,"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\/1027744"}],"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=1027744"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1027744\/revisions"}],"predecessor-version":[{"id":1027764,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1027744\/revisions\/1027764"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/1027760"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=1027744"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=1027744"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=1027744"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}