\u4e0d\u662f\u5f88\u559c\u6b22<\/strong>\u8fd9\u53e5\uff0cjieba\u4f1a\u5206\u6210\u2019\u4e0d\u662f\u2018\uff0c\u2019\u5f88\u559c\u6b22\u2018\u4e24\u4e2a\u8bcd\uff0c\u8fd9\u6837\u5bfc\u81f4\u8fd9\u53e5\u8bdd\u5f88\u5927\u6982\u7387\u4f1a\u88ab\u9884\u6d4b\u4e3a\u597d\u8bc4\u3002<\/p>\n\u6240\u4ee5\u8fd9\u91cc\u6211\u5728\u81ea\u5b9a\u4e49\u7684\u8bcd\u5178\u4e2d\u5206\u597d\u4e86\u5f88\u591a\u7c7b\u4f3c\u8fd9\u6837\u7684\u8bcd\uff0c\u63d0\u9ad8\u4e86\u4e00\u70b9\u70b9\u51c6\u786e\u7387\u3002<\/p>\n
\u7136\u540e\u5c06\u5168\u90e8\u7684\u8bed\u6599\u63091:4\u5206\u4e3a\u6d4b\u8bd5\u96c6\u4e0e\u8bad\u7ec3\u96c6<\/p>\n
n = len(review_list) \/\/ 5\n\ntrain_review_list, train_sentiment_list = review_list[n:], sentiment_list[n:]\ntest_review_list, test_sentiment_list = review_list[:n], sentiment_list[:n]<\/code><\/pre>\n\u5206\u8bcd<\/strong><\/p>\n\u4f7f\u7528jieba\u5206\u8bcd\uff0c\u5c06\u8bed\u6599\u8fdb\u884c\u5206\u8bcd\uff0c\u5e76\u4e14\u53bb\u9664stopwords\u3002<\/p>\n
import re\nimport jieba\n\nstopword_path = '.\/data\/stopwords.txt'\n\ndef load_stopwords(file_path):\n stop_words = []\n with open(file_path, encoding='UTF-8') as words:\n stop_words.extend([i.strip() for i in words.readlines()])\n return stop_words\n\ndef review_to_text(review):\n stop_words = load_stopwords(stopword_path)\n # \u53bb\u9664\u82f1\u6587\n review = re.sub("[^\\u4e00-\\u9fa5^a-z^A-Z]", '', review)\n review = jieba.cut(review)\n # \u53bb\u6389\u505c\u7528\u8bcd\n if stop_words:\n all_stop_words = set(stop_words)\n words = [w for w in review if w not in all_stop_words]\n\n return words\n\n# \u7528\u4e8e\u8bad\u7ec3\u7684\u8bc4\u8bba\nreview_train = [' '.join(review_to_text(review)) for review in train_review_list]\n# \u5bf9\u4e8e\u8bad\u7ec3\u8bc4\u8bba\u5bf9\u5e94\u7684\u597d\u8bc4\/\u5dee\u8bc4\nsentiment_train = train_sentiment_list\n\n# \u7528\u4e8e\u6d4b\u8bd5\u7684\u8bc4\u8bba\nreview_test = [' '.join(review_to_text(review)) for review in test_review_list]\n# \u5bf9\u4e8e\u6d4b\u8bd5\u8bc4\u8bba\u5bf9\u5e94\u7684\u597d\u8bc4\/\u5dee\u8bc4\nsentiment_test = test_sentiment_list<\/code><\/pre>\nTF*IDF\u4e0e\u8bcd\u9891\u5411\u91cf\u5316<\/strong><\/p>\nTF-IDF\uff08\u662f\u4e00\u79cd\u5e38\u7528\u4e8e\u4fe1\u606f\u5904\u7406\u548c\u6570\u636e\u6316\u6398\u7684\u52a0\u6743\u6280\u672f\u3002\u6839\u636e\u8bcd\u8bed\u7684\u5728\u6587\u672c\u4e2d\u51fa\u73b0\u7684\u6b21\u6570\u548c\u5728\u6574\u4e2a\u8bed\u6599\u4e2d\u51fa\u73b0\u7684\u6587\u6863\u9891\u7387\u6765\u8ba1\u7b97\u4e00\u4e2a\u8bcd\u8bed\u5728\u6574\u4e2a\u8bed\u6599\u4e2d\u7684\u91cd\u8981\u7a0b\u5ea6\u3002<\/p>\n
\u5b83\u7684\u4f18\u70b9\u662f\u80fd\u8fc7\u6ee4\u6389\u4e00\u4e9b\u5e38\u89c1\u7684\u5374\u65e0\u5173\u7d27\u8981\u672c\u7684\u8bcd\u8bed\uff0c\u540c\u65f6\u4fdd\u7559\u5f71\u54cd\u6574\u4e2a\u6587\u672c\u7684\u91cd\u8981\u5b57\u8bcd\u3002<\/p>\n
\u4f7f\u7528Countvectorizer()<\/code>\u5c06\u4e00\u4e2a\u6587\u6863\u8f6c\u6362\u4e3a\u5411\u91cf\uff0c\u8ba1\u7b97\u8bcd\u6c47\u5728\u6587\u672c\u4e2d\u51fa\u73b0\u7684\u9891\u7387\u3002<\/p>\nCountVectorizer<\/code>\u7c7b\u4f1a\u5c06\u6587\u672c\u4e2d\u7684\u8bcd\u8bed\u8f6c\u6362\u4e3a\u8bcd\u9891\u77e9\u9635\uff0c\u4f8b\u5982\u77e9\u9635\u4e2d\u5305\u542b\u4e00\u4e2a\u5143\u7d20a[i] [j]\uff0c\u5b83\u8868\u793aj\u8bcd\u5728i\u7c7b\u6587\u672c\u4e0b\u7684\u8bcd\u9891\u3002\u5b83\u901a\u8fc7fit_transform\u51fd\u6570\u8ba1\u7b97\u5404\u4e2a\u8bcd\u8bed\u51fa\u73b0\u7684\u6b21\u6570\u3002<\/p>\nTfidfTransformer\u7528\u4e8e\u7edf\u8ba1vectorizer\u4e2d\u6bcf\u4e2a\u8bcd\u8bed\u7684TF-IDF\u503c\u3002<\/p>\n
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.naive_bayes import MultinomialNB\n\ncount_vec = CountVectorizer(max_df=0.8, min_df=3)\n\ntfidf_vec = TfidfVectorizer()\n\n# \u5b9a\u4e49Pipeline\u5bf9\u5168\u90e8\u6b65\u9aa4\u7684\u6d41\u5f0f\u5316\u5c01\u88c5\u548c\u7ba1\u7406\uff0c\u53ef\u4ee5\u5f88\u65b9\u4fbf\u5730\u4f7f\u53c2\u6570\u96c6\u5728\u65b0\u6570\u636e\u96c6\uff08\u6bd4\u5982\u6d4b\u8bd5\u96c6\uff09\u4e0a\u88ab\u91cd\u590d\u4f7f\u7528\u3002\ndef MNB_Classifier():\n return Pipeline([\n ('count_vec', CountVectorizer()),\n ('mnb', MultinomialNB())\n ])<\/code><\/pre>\nmax_df<\/code> \u8fd9\u4e2a\u53c2\u6570\u7684\u4f5c\u7528\u662f\u4f5c\u4e3a\u4e00\u4e2a\u9608\u503c\uff0c\u5f53\u6784\u9020\u8bed\u6599\u5e93\u7684\u5173\u952e\u8bcd\u96c6\u7684\u65f6\u5019\uff0c\u5982\u679c\u67d0\u4e2a\u8bcd\u7684\u8bcd\u9891\u5927\u4e8emax_df<\/code>\uff0c\u8fd9\u4e2a\u8bcd\u4e0d\u4f1a\u88ab\u5f53\u4f5c\u5173\u952e\u8bcd\u3002<\/p>\n\u5982\u679c\u8fd9\u4e2a\u53c2\u6570\u662ffloat\uff0c\u5219\u8868\u793a\u8bcd\u51fa\u73b0\u7684\u6b21\u6570\u4e0e\u8bed\u6599\u5e93\u6587\u6863\u6570\u7684\u767e\u5206\u6bd4\uff0c\u5982\u679c\u662fint\uff0c\u5219\u8868\u793a\u8bcd\u51fa\u73b0\u7684\u6b21\u6570\u3002<\/p>\n
min_df<\/code>\u7c7b\u4f3c\u4e8emax_df<\/code>\uff0c\u4e0d\u540c\u4e4b\u5904\u5728\u4e8e\u5982\u679c\u67d0\u4e2a\u8bcd\u7684\u8bcd\u9891\u5c0f\u4e8emin_df<\/code>\uff0c\u5219\u8fd9\u4e2a\u8bcd\u4e0d\u4f1a\u88ab\u5f53\u4f5c\u5173\u952e\u8bcd<\/p>\n\u8fd9\u6837\u6211\u4eec\u5c31\u6210\u529f\u7684\u6784\u9020\u51fa\u4e86\u7528\u4e8e\u8bad\u7ec3\u4ee5\u53ca\u6d4b\u8bd5\u7684Pipeline<\/p>\n
\u7136\u540e\u7528 Pipeline.fit<\/strong>()\u5bf9\u8bad\u7ec3\u96c6\u8fdb\u884c\u8bad\u7ec3<\/p>\n\u518d\u76f4\u63a5\u7528 Pipeline.score()<\/strong> \u5bf9\u6d4b\u8bd5\u96c6\u8fdb\u884c\u9884\u6d4b\u5e76\u8bc4\u5206<\/p>\nmnbc_clf = MNB_Classifier()\n\n# \u8fdb\u884c\u8bad\u7ec3\nmnbc_clf.fit(review_train, sentiment_train)\n\n# \u6d4b\u8bd5\u96c6\u51c6\u786e\u7387\nprint('\u6d4b\u8bd5\u96c6\u51c6\u786e\u7387\uff1a {}'.format(mnbc_clf.score(review_test, sentiment_test)))<\/code><\/pre>\n\u8fd9\u6837\u6211\u4eec\u5c31\u5b8c\u6210\u4e86\u6574\u4e2a\u4ece\u8bad\u7ec3\u5230\u6d4b\u8bd5\u7684\u5168\u90e8\u6d41\u7a0b\u3002<\/p>\n
\u57fa\u672c\u4e0a\u6d4b\u8bd5\u96c6\u7684\u6b63\u786e\u7387\u572879%-80%\u5de6\u53f3\u3002<\/p>\n
\u56e0\u4e3a\u7535\u5f71\u8bc4\u8bba\u4e2d\u6709\u5f88\u5927\u4e00\u90e8\u5206\u597d\u8bc4\u4e2d\u4f1a\u6709\u8d1f\u9762\u60c5\u611f\u7684\u8bcd\u8bed\uff0c\u4f8b\u5982\u5728\u7eaa\u5f55\u7247\u300a\u6d77\u8c5a\u6e7e\u300b\u4e2d<\/p>\n
\n\u6211\u89c9\u5f97\u5927\u90e8\u5206\u770b\u672c\u7247\u4f1a\u6709\u611f\u7684\u4eba\uff0c\u90fd\u4e0d\u77e5\u9053\uff0c\u4e2d\u56fd\u7684\u767d\u66a8\u8c5a\u5df2\u7ecf\u706d\u7edd8\u5e74\u4e86\uff0c\u4e5f\u4e0d\u4f1a\u77e5\u9053\uff0c\u957f\u6c5f\u91cc\u7684\u6c5f\u8c5a\u4e5f\u4ec5\u52691000\u5de6\u53f3\u4e86\u3002\u4e0e\u5176\u611f\u6168\uff0c\u5492\u9a82\u65e5\u672c\u4eba\u5982\u4f55\u6355\u6740\u6d77\u8c5a\uff0c\u4e0d\u5982\u505a\u4e9b\u5b9e\u9645\u7684\u4e8b\u60c5\uff0c\u4fdd\u62a4\u4e00\u4e0b\u957f\u6c5f\u91cc\u7684\u6c5f\u8c5a\u5427\uff0c\u6ca1\u51e0\u5e74\uff0c\u4e5f\u5c06\u7edd\u8ff9\u4e86\u3002\u4e2d\u56fd\u4eba\u505a\u51fa\u6765\u7684\u4e8b\u60c5\uff0c\u4e5f\u4e0d\u4f1a\u6bd4\u5c0f\u65e5\u672c\u597d\u5230\u54ea\u513f\u53bb\u3002<\/p>\n<\/blockquote>\n
\u6240\u4ee5\u8bf4\u5982\u679c\u5c06\u8fd9\u79cd\u7c7b\u4f3c\u7684\u597d\u8bc4\u53bb\u9664\uff0c\u5219\u53ef\u4ee5\u63d0\u9ad8\u51c6\u786e\u7387\u3002<\/p>\n
\u4fdd\u5b58\u8bad\u7ec3\u597d\u7684\u6a21\u578b<\/strong><\/p>\n# \u5148\u8f6c\u6362\u6210\u8bcd\u9891\u77e9\u9635\uff0c\u518d\u8ba1\u7b97TFIDF\u503c\ntfidf = tfidftransformer.fit_transform(vectorizer.fit_transform(review_train))\n# \u6734\u7d20\u8d1d\u53f6\u65af\u4e2d\u7684\u591a\u9879\u5f0f\u5206\u7c7b\u5668\nclf = MultinomialNB().fit(tfidf, sentiment_train)\n\nwith open(model_export_path, 'wb') as file:\n d = {\n "clf": clf,\n "vectorizer": vectorizer,\n "tfidftransformer": tfidftransformer,\n }\n pickle.dump(d, file)<\/code><\/pre>\n\u4f7f\u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u8fdb\u884c\u5f71\u8bc4\u60c5\u611f\u9884\u6d4b<\/strong><\/p>\n\u8fd9\u91cc\u6211\u76f4\u63a5\u8d34\u4e0a\u5168\u90e8\u7684\u6e90\u4ee3\u7801\uff0c\u4ee3\u7801\u975e\u5e38\u7b80\u5355\uff0c\u6211\u5c06\u6574\u4e2a\u5904\u7406\u903b\u8f91\u5c01\u88c5\u4e3a\u4e00\u4e2a\u7c7b\uff0c\u8fd9\u6837\u5c31\u975e\u5e38\u65b9\u4fbf\u4f7f\u7528\u4e86\u3002<\/p>\n
\u6709\u9700\u8981\u76f4\u63a5\u53ef\u4ee5\u5728\u6211\u7684Github\u4e0aclone<\/p>\n
# -*- coding: utf-8 -*-\nimport re\nimport pickle\n\nimport numpy as np\nimport jieba\n\nclass SentimentAnalyzer(object):\n def __init__(self, model_path, userdict_path, stopword_path):\n self.clf = None\n self.vectorizer = None\n self.tfidftransformer = None\n self.model_path = model_path\n self.stopword_path = stopword_path\n self.userdict_path = userdict_path\n self.stop_words = []\n self.tokenizer = jieba.Tokenizer()\n self.initialize()\n\n # \u52a0\u8f7d\u6a21\u578b\n def initialize(self):\n with open(self.stopword_path, encoding='UTF-8') as words:\n self.stop_words = [i.strip() for i in words.readlines()]\n\n with open(self.model_path, 'rb') as file:\n model = pickle.load(file)\n self.clf = model['clf']\n self.vectorizer = model['vectorizer']\n self.tfidftransformer = model['tfidftransformer']\n if self.userdict_path:\n self.tokenizer.load_userdict(self.userdict_path)\n\n # \u8fc7\u6ee4\u6587\u5b57\u4e2d\u7684\u82f1\u6587\u4e0e\u65e0\u5173\u6587\u5b57\n def replace_text(self, text):\n text = re.sub('((https?|ftp|file):\/\/)?[-A-Za-z0-9+&@#\/%?=~_|!:,.;]+[-A-Za-z0-9+&@#\/%=~_|].(com|cn)', '', text)\n text = text.replace('\\u3000', '').replace('\\xa0', '').replace('\u201d', '').replace('"', '')\n text = text.replace(' ', '').replace('\u21b5', '').replace('\\n', '').replace('\\r', '').replace('\\t', '').replace('\uff09', '')\n text_corpus = re.split('[\uff01\u3002\uff1f\uff1b\u2026\u2026;]', text)\n return text_corpus\n\n # \u60c5\u611f\u5206\u6790\u8ba1\u7b97\n def predict_score(self, text_corpus):\n # \u5206\u8bcd\n docs = [self.__cut_word(sentence) for sentence in text_corpus]\n new_tfidf = self.tfidftransformer.transform(self.vectorizer.transform(docs))\n predicted = self.clf.predict_proba(new_tfidf)\n # \u56db\u820d\u4e94\u5165\uff0c\u4fdd\u7559\u4e09\u4f4d\n result = np.around(predicted, decimals=3)\n return result\n\n # jieba\u5206\u8bcd\n def __cut_word(self, sentence):\n words = [i for i in self.tokenizer.cut(sentence) if i not in self.stop_words]\n result = ' '.join(words)\n return result\n\n def analyze(self, text):\n text_corpus = self.replace_text(text)\n result = self.predict_score(text_corpus)\n\n neg = result[0][0]\n pos = result[0][1]\n\n print('\u5dee\u8bc4\uff1a {} \u597d\u8bc4\uff1a {}'.format(neg, pos))<\/code><\/pre>\n\u4f7f\u7528\u65f6\u53ea\u8981\u5b9e\u4f8b\u5316\u8fd9\u4e2a\u5206\u6790\u5668\uff0c\u5e76\u4f7f\u7528analyze()<\/code>\u65b9\u6cd5\u5c31\u53ef\u4ee5\u4e86\u3002<\/p>\n# -*- coding: utf-8 -*-\nfrom native_bayes_sentiment_analyzer import SentimentAnalyzer\n\nmodel_path = '.\/data\/bayes.pkl'\nuserdict_path = '.\/data\/userdict.txt'\nstopword_path = '.\/data\/stopwords.txt'\ncorpus_path = '.\/data\/review.csv'\n\nanalyzer = SentimentAnalyzer(model_path=model_path, stopword_path=stopword_path, userdict_path=userdict_path)\ntext = '\u500d\u611f\u5931\u671b\u7684\u4e00\u90e8\u8bfa\u5170\u7684\u7535\u5f71\uff0c\u611f\u89c9\u66f4\u50cf\u662f\u76d7\u68a6\u5e2e\u7684\u4e00\u573a\u5927\u6742\u70e9\u3002\u867d\u7136\u770b\u4e4b\u524d\u5c31\u77e5\u9053\u80af\u5b9a\u662f\u4e00\u90e8\u65e0\u6cd5\u8d85\u8d8a\u524d\u4f202\u7684\u8759\u8760\u72ed\uff0c\u4f46\u771f\u5fc3\u6ca1\u60f3\u5230\u80fd\u5dee\u5230\u8fd9\u4e2a\u5730\u6b65\u3002\u8282\u594f\u7684\u628a\u63a7\u7684\u5931\u8bef\u548c\u89d2\u8272\u7684\u5b9a\u4f4d\u6a21\u7cca\u7edd\u5bf9\u662f\u6574\u90e8\u5f71\u7247\u7684\u786c\u4f24\u3002'\nanalyzer.analyze(text=text)<\/code><\/pre>\ngithub.com\/3inchtime\/d\u2026<\/a><\/p>\n\u4ee5\u4e0a\u5168\u90e8\u4ee3\u7801\u5747push\u5230\u4e86\u6211\u7684Github\u4e0a\uff0c\u6b22\u8fce\u5927\u5bb6\u63d0\u51fa\u5efa\u8bae\u3002<\/p>\n
\u4f5c\u8005\uff1a3inchtime
\n\u94fe\u63a5\uff1ahttps:\/\/juejin.cn\/post\/6844903941226921991<\/a>
\n\u6765\u6e90\uff1a\u7a00\u571f\u6398\u91d1
\n\u8457\u4f5c\u6743\u5f52\u4f5c\u8005\u6240\u6709\u3002\u5546\u4e1a\u8f6c\u8f7d\u8bf7\u8054\u7cfb\u4f5c\u8005\u83b7\u5f97\u6388\u6743\uff0c\u975e\u5546\u4e1a\u8f6c\u8f7d\u8bf7\u6ce8\u660e\u51fa\u5904\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"\u524d\u8a00 \u672c\u6587\u4f7f\u7528\u6734\u7d20\u8d1d\u53f6\u65af\u7b97\u6cd5\u5b9e\u73b0 \u8c46\u74e3Top250\u7535\u5f71\u8bc4\u4ef7\u7684\u60c5\u611f\u5206\u6790\u4e0e\u9884\u6d4b\u3002 \u6700\u8fd1\u5728\u5b66\u4e60\u81ea\u7136\u8bed\u8a00\u6b63\u8d1f\u9762\u60c5\u611f\u7684\u5904\u7406\u95ee\u9898\uff0c\u4f46\u662f\u7edd\u5927\u90e8\u5206\u80fd\u641c\u7d22 \u2026 …<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0},"categories":[59],"tags":[],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9424"}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9424"}],"version-history":[{"count":1,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9424\/revisions"}],"predecessor-version":[{"id":9425,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9424\/revisions\/9425"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9424"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9424"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9424"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"image-20250610201617997\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20250610201617997.png\")
<\/p>\n![\"image-20250610201632595\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20250610201632595.png\")
<\/p>\n![\"image-20250610201647365\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20250610201647365.png\")
<\/p>\n