📖 第13课：情感分析

from sklearn.feature_extraction.text import TfidfVectorizer from sklearn.naive_bayes import MultinomialNB from sklearn.model_selection import train_test_split from sklearn.metrics import classification_report import numpy as np # 模拟中文情感数据集 texts = [ "这部电影非常好看，剧情精彩", "太棒了，强烈推荐", "令人感动的故事", "演员演技出色，值得观看", "音乐优美，画面精致", "经典之作，百看不厌", "非常满意，超出预期", "温馨治愈，心情愉悦", "精彩的表演，完美的结局", "好看得停不下来", "质量上乘，诚意之作", "故事引人入胜", "这部电影太差了，浪费时间", "剧情无聊，看不下去", "演技尴尬，令人失望", "浪费钱，不推荐", "太烂了，简直受不了", "特效廉价，逻辑混乱", "无聊至极，毫无亮点", "导演水平堪忧", "烂片无疑，别去看", "失望透顶，后悔购票", "毫无新意，套路满满", "演技浮夸，台词尴尬", "还行吧，一般般", "有点无聊但能看", "中规中矩，没什么惊喜", "不好不坏，凑合", "部分场景不错，整体一般", "有待提高", "有些亮点但整体平淡", "比预期差一点", ] labels = (['positive'] * 12 + ['negative'] * 12 + ['neutral'] * 8) # TF-IDF + 朴素贝叶斯 vectorizer = TfidfVectorizer(token_pattern=r'(?u)\b\w+\b', max_features=100) X = vectorizer.fit_transform(texts) y = np.array(labels) X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42, stratify=y) clf = MultinomialNB(alpha=1.0) clf.fit(X_train, y_train) y_pred = clf.predict(X_test) print("=== 情感分析（TF-IDF + 朴素贝叶斯） ===") print(classification_report(y_test, y_pred)) # 对新文本预测 new_texts = ["这部电影真的很好看", "太难看了，浪费生命", "还行吧一般般"] new_X = vectorizer.transform(new_texts) predictions = clf.predict(new_X) probas = clf.predict_proba(new_X) for text, pred, proba in zip(new_texts, predictions, probas): print(f" '{text}' → {pred} (概率: {dict(zip(clf.classes_, proba.round(3)))})")

from sklearn.feature_extraction.text import TfidfVectorizer from sklearn.linear_model import LogisticRegression from sklearn.svm import LinearSVC from sklearn.ensemble import RandomForestClassifier from sklearn.model_selection import cross_val_score import numpy as np texts = [ "这部电影非常好看剧情精彩", "太棒了强烈推荐", "令人感动的故事", "演员演技出色值得观看", "音乐优美画面精致", "经典之作百看不厌", "非常满意超出预期", "温馨治愈心情愉悦", "这部电影太差了浪费时间", "剧情无聊看不下去", "演技尴尬令人失望", "浪费钱不推荐", "太烂了简直受不了", "特效廉价逻辑混乱", "无聊至极毫无亮点", "烂片无疑别去看", "还行吧一般般", "有点无聊但能看", "中规中矩没什么惊喜", "不好不坏凑合", ] labels = ['pos'] * 8 + ['neg'] * 8 + ['neu'] * 4 vectorizer = TfidfVectorizer(token_pattern=r'(?u)\b\w+\b', max_features=200) X = vectorizer.fit_transform(texts) y = np.array(labels) # 模型对比 models = { '朴素贝叶斯': MultinomialNB(), '逻辑回归': LogisticRegression(max_iter=1000, random_state=42), 'SVM': LinearSVC(random_state=42, max_iter=2000), '随机森林': RandomForestClassifier(n_estimators=50, random_state=42), } print("=== 情感分析模型对比（5折交叉验证） ===") for name, model in models: pass # just structure for name, model in models.items(): try: scores = cross_val_score(model, X, y, cv=3, scoring='accuracy') print(f" {name:10s}: 准确率={scores.mean():.4f} (±{scores.std():.4f})") except Exception as e: print(f" {name:10s}: 错误 - {e}") # 情感词典方法 positive_words = {'好', '棒', '精彩', '感动', '出色', '优美', '经典', '满意', '温馨', '治愈', '推荐'} negative_words = {'差', '无聊', '尴尬', '失望', '浪费', '烂', '廉价', '混乱', '至极', '毫无', '堪忧', '后悔'} def lexicon_sentiment(text): pos_count = sum(1 for w in positive_words if w in text) neg_count = sum(1 for w in negative_words if w in text) if pos_count > neg_count: return 'positive' elif neg_count > pos_count: return 'negative' return 'neutral' print("\n=== 情感词典方法 ===") test_texts = ["这部电影好看推荐", "太差了浪费时间", "一般般没有亮点"] for text in test_texts: result = lexicon_sentiment(text) print(f" '{text}' → {result}")