Duplicate from EugeneMeng/Hongguo-Short-Drama-Corpus-AI-Labeled

Co-authored-by: Eugene Meng <EugeneMeng@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,60 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,66 @@
+---
+license: cc-by-nc-sa-4.0
+task_categories:
+- text-classification
+- feature-extraction
+language:
+- zh
+tags:
+- short-drama
+- entertainment
+- marketing
+size_categories:
+- 1K<n<10K
+---
+
+# 🎬 2025年红果短剧全量语料库 (AI标注版 V1)
+## Hongguo Short Drama Corpus with AI-Augmented Audience Labels
+
+### 1. 数据集简介 (Dataset Summary)
+本数据集包含了约 1500 条来自红果 (Hongguo/RedFruit) 平台的微短剧精选数据。本数据集旨在为中文短剧的 NLP 研究、市场趋势分析以及自动化剧名生成等任务提供高质量的基准数据。
+
+**核心特色：**
+- **多维度标注**：涵盖了标题、受众、标签、简介及集数。
+- **AI 增强受众标签**：针对部分原始数据未标注受众标签的问题，使用了专门的 `sex_divide.py` ，利用进行“TF-IDF + 逻辑回归”预测，并保留了预测置信度。
+
+### 2. 数据字段说明 (Data Fields)
+
+| 字段名 | 类型 | 说明 |
+| :--- | :--- | :--- |
+| `drama_id` | string | 脱敏后的剧集唯一编号 (例如 drama_0001) |
+| `title` | string | 短剧标题 |
+| `audience_type` | string | 受众类型：男频 / 女频 |
+| `tags` | list/string | 剧集标签 (如：都市, 重生, 逆袭) |
+| `episode_count` | int | 剧集总集数 |
+| `description` | string | 剧情详细简介 |
+| `ai_confidence` | float | AI 预测受众标签的置信度 (0.0 - 1.0) |
+| `label_source` | string | 受众数据来源：platform (原厂) / model_prediction (AI预测) |
+
+### 3. 数据来源与标注逻辑 (Methodology)
+- **原始数据**：采集自红果平台公开信息。
+- **标注过程**：
+    1. 提取平台原有的受众标签。
+    2. 对于缺失标签的样本，通过 `sex_divide.py` 脚本，基于标题和简介的关键词、语义特征进行性别偏好预测。
+    3. 预测逻辑已开源在数据集的 `scripts/` 目录下。
+
+### 4. 如何使用 (Usage)
+你可以直接使用 Hugging Face 的 `datasets` 库或 `pandas` 加载：
+
+```python
+import pandas as pd
+
+# 加载数据
+df = pd.read_parquet("short_drama_corpus_v1.parquet")
+print(df.head())
+```
+
+
+
+**⚠️ 声明**    
+-数据属性 ：本数据集中的“受众类型”字段包含由“TF-IDF + 逻辑回归”(sex_divide.py) 预测的结果。预测结果仅供学术参考，不代表 100% 的真实市场分布，亦不代表原平台的官方分类。  
+-合规性：本数据集由个人出于学术兴趣采集并清洗。用户在下载和使用过程中，请务必遵守原平台的服务条款。因不当使用该数据而产生的任何法律纠纷，由使用者自行承担，原作者概不负责。  
+-禁止商业化 ：本数据集严禁用于任何形式的商业盈利活动。  
+**⚠️ Disclaimer**    
+-Data Accuracy: The audience_type field in this dataset contains results predicted by an AI model (sex_divide.py). These predictions are for academic reference only and do not represent the absolute real-world distribution or the official classification of the original platform.  
+-Compliance: This dataset was collected and cleaned by individuals for academic interests. Users must comply with the Terms of Service of the original platform. The author is not responsible for any legal disputes arising from improper use.  
+-Non-commercial: Commercial use of this dataset is strictly prohibited.  

scripts/sex_divide.py

@@ -0,0 +1,154 @@
+import pandas as pd
+import numpy as np
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.linear_model import LogisticRegression
+from sklearn.model_selection import cross_val_score, StratifiedKFold
+from sklearn.pipeline import Pipeline
+import os
+
+# =================配置区域=================
+# 输入文件路径 
+input_path = r""
+# 输出文件路径 (自动生成在同目录下，文件名加后缀)
+output_path = input_path.replace(".xlsx", "_AI全标注版.xlsx")
+
+# 检查是否有xlsxwriter模块
+try:
+    import xlsxwriter
+    HAS_XLSXWRITER = True
+except ImportError:
+    HAS_XLSXWRITER = False
+
+
+# =========================================
+
+def main():
+    print(f"正在读取文件: {input_path} ...")
+    try:
+        df = pd.read_excel(input_path)
+    except FileNotFoundError:
+        print(f"错误：找不到文件，请检查路径是否正确：{input_path}")
+        return
+
+    # 1. 数据预处理
+    # 构造文本特征：将标题、标签、简介合并，能最大化利用信息
+    # fillna('') 防止空值报错
+    df['combined_text'] = df['标题'].fillna('') + ' ' + df['标签'].fillna('') + ' ' + df['简介'].fillna('')
+
+    # 拆分数据：有标签的（训练集） vs 没标签的（待预测集）
+    # 假设已有标签列名为 '受众'，且值为 '男频' 或 '女频'
+    train_mask = df['受众'].isin(['男频', '女频'])
+    train_df = df[train_mask].copy()
+    predict_df = df[~train_mask].copy()
+
+    print(f"数据加载成功！")
+    print(f"   - 训练集(已标注): {len(train_df)} 条")
+    print(f"   - 待预测(未标注): {len(predict_df)} 条")
+
+    if len(train_df) == 0:
+        print("错误：未找到已标注为'男频'或'女频'的数据，请检查'受众'列。")
+        return
+
+    # 2. 构建模型管道 (TF-IDF + 逻辑回归)
+    # analyzer='char': 按字切分，对中文短文本效果好，无需jieba分词
+    # ngram_range=(1, 3): 提取1到3个字的组合特征（如"战","战神","系统"）
+    pipeline = Pipeline([
+        ('tfidf', TfidfVectorizer(analyzer='char', ngram_range=(1, 3), max_features=5000)),
+        ('clf', LogisticRegression(class_weight='balanced', solver='liblinear', random_state=42))
+    ])
+
+    # 3. 准备训练标签 (映射为 0 和 1)
+    label_map = {'男频': 0, '女频': 1}
+    inverse_map = {0: '男频', 1: '女频'}
+    y_train = train_df['受众'].map(label_map)
+
+    # 4. 学术验证：计算交叉验证准确率
+    print("\n进行5折交叉验证 (用于学术支撑)...")
+    cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)
+    scores = cross_val_score(pipeline, train_df['combined_text'], y_train, cv=cv, scoring='accuracy')
+    print(
+        f"模型平均准确率: {scores.mean() * 100:.2f}% (Range: {scores.min() * 100:.2f}% - {scores.max() * 100:.2f}%)")
+    print("   -> (提示：如果准确率超过90%，说明该技术路线在学术上非常可靠)")
+
+    # 5. 全量训练
+    print("\n正在全量训练并预测...")
+    pipeline.fit(train_df['combined_text'], y_train)
+
+    # 6. 预测未标注数据
+    if len(predict_df) > 0:
+        # 预测类别
+        preds = pipeline.predict(predict_df['combined_text'])
+        # 预测概率（用于置信度）
+        probs = pipeline.predict_proba(predict_df['combined_text'])
+
+        # 将结果填回 DataFrame
+        predict_df['受众'] = [inverse_map[p] for p in preds]  # 填补缺失的标签
+        predict_df['AI_预测置信度'] = np.max(probs, axis=1)  # 取最大概率值
+
+        # 给训练集也加上置信度（虽然是100%确定的，但也补全格式）
+        # 这里我们填 1.0 或者填模型对训练集的预测概率
+        train_probs = pipeline.predict_proba(train_df['combined_text'])
+        train_df['AI_预测置信度'] = np.max(train_probs, axis=1)
+
+        # 标记数据来源，方便您后续区分
+        train_df['数据来源'] = '人工标注(原有)'
+        predict_df['数据来源'] = 'AI预测'
+
+        # 合并结果
+        final_df = pd.concat([train_df, predict_df])
+    else:
+        final_df = train_df
+        print("没有需要预测的数据。")
+
+    # 7. 导出特征重要性 (用于论文解释)
+    feature_names = pipeline.named_steps['tfidf'].get_feature_names_out()
+    coefs = pipeline.named_steps['clf'].coef_[0]
+    top_n = 10
+
+    # 排序
+    top_female_idx = np.argsort(coefs)[::-1][:top_n]  # 权重最大的正值 (女频)
+    top_male_idx = np.argsort(coefs)[:top_n]  # 权重最小的负值 (男频)
+
+    print("\n最具区分度的特征词 (Feature Importance):")
+    print(f"女频特征: {[feature_names[i] for i in top_female_idx]}")
+    print(f"男频特征: {[feature_names[i] for i in top_male_idx]}")
+
+    # 8. 保存文件
+    # 删除临时辅助列 'combined_text'
+    if 'combined_text' in final_df.columns:
+        del final_df['combined_text']
+
+    # 解决ID列变成科学计数法的问题
+    # 方法1：将ID列转换为���符串类型
+    id_columns = ['id', 'ID', '编号', '序号']  # 常见的ID列名
+    for col in id_columns:
+        if col in final_df.columns:
+            final_df[col] = final_df[col].astype(str)
+            print(f"已将 '{col}' 列转换为文本格式，避免科学计数法显示")
+            break
+    
+    # 方法2：使用xlsxwriter进行精确格式控制
+    if HAS_XLSXWRITER:
+        print("🔧 使用xlsxwriter引擎进行精确格式控制...")
+        with pd.ExcelWriter(output_path, engine='xlsxwriter') as writer:
+            final_df.to_excel(writer, index=False, sheet_name='AI标注结果')
+            
+            # 获取工作表对象
+            worksheet = writer.sheets['AI标注结果']
+            
+            # 设置ID列格式为文本
+            for col_idx, col_name in enumerate(final_df.columns):
+                if col_name.lower() in ['id', '编号', '序号']:
+                    # 设置该列为文本格式
+                    worksheet.set_column(col_idx, col_idx, 20, None, {'num_format': '@'})
+                    print(f"已设置 '{col_name}' 列为文本格式")
+                    break
+    else:
+        # 如果没有xlsxwriter，使用默认方法
+        print("未安装xlsxwriter，使用基础方法保存...")
+        final_df.to_excel(output_path, index=False)
+    print(f"\n处理完成！文件已保存至:\n   {output_path}")
+
+
+if __name__ == "__main__":
+    main()

short_drama_corpus_v1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3770713fb27b4d98fa759f15894564ca64bcdd93bc2d7de0f45c7f65fed79429
+size 598137