意图识别

• 意图识别DEMO

      
    import pandas as pd
    import jieba
    from sklearn.feature_extraction.text import TfidfVectorizer
    from sklearn.linear_model import LogisticRegression
    from sklearn.pipeline import Pipeline
    from sklearn.metrics import classification_report, accuracy_score
    from sklearn.model_selection import train_test_split
      
    # 示例数据（假设已经有了标注好的意图数据）
    data = {
        'text': ["订一张去北京的机票",
                 "给我看看最新的智能手机型号",
                 "明天的天气怎么样？",
                 "你觉的天气怎么样？",
                  "python怎么样？",
                 "告诉我关于Python编程的信息"],
        'intent': ["flight_booking", 
                   "product_information", 
                   "weather_information", 
                   "weather_information",
                   "software_information",
                   "software_information"]
    }
      
    df = pd.DataFrame(data)
      
    # 分词函数（使用jieba分词）
    def chinese_tokenizer(text): 
        return list(jieba.cut(text))
      
    # 划分训练集和测试集
    train_texts, test_texts, train_labels, test_labels = train_test_split(df['text'], df['intent'], test_size=0.2, random_state=42)
      
    # 定义TF-IDF向量化器和分类器（这里使用逻辑回归作为分类器），注意不指定token_pattern参数
    tfidf_vectorizer = TfidfVectorizer(tokenizer=chinese_tokenizer)
    classifier = LogisticRegression(max_iter=1000)
      
    # 构建管道
    model = Pipeline([
        ('tfidf', tfidf_vectorizer),
        ('clf', classifier),
    ])
      
    # 训练模型
    model.fit(train_texts, train_labels)
      
    # 在测试集上评估模型
    predictions = model.predict(test_texts)
      
    print("Accuracy:", accuracy_score(test_labels, predictions))
    print("\nClassification Report:")
    print(classification_report(test_labels, predictions))
      
    # 准备测试数据
    test_data = ["预定一间去上海的酒店",
                 "我如何学习机器学习？",
                 "今天北京的温度是多少？"]
      
    # 预测测试数据的意图
    predictions = model.predict(test_data)
      
    # 输出预测结果
    for text, intent in zip(test_data, predictions):
        print(f"Text: {text} | Predicted Intent: {intent}")