AdsTrace: A Multimodal Benchmark Dataset for Second-level CTR and Advertising Effectiveness Prediction

[Paper Link (Coming Soon)] | [Project Repository] | [Dataset DOI (via Zenodo/HF)]

English | 中文说明

---

🇬🇧 English Description

🌟 Key Features

AdsTrace is a novel, large-scale multimodal benchmark dataset derived from real-world e-commerce short-video advertising. It bridges the gap between fine-grained multimodal content (visual, acoustic, textual) and actual business outcomes.

🌟 Dataset Highlights

• High-Resolution Labels: Provides second-level instantaneous Click-Through Rate (iCTR) trajectories, moving beyond static "black-box" video analysis.
• Verified Business Metrics: Every video is paired with validated financial outcomes, including Pay ROI and Conversion Rate (CVR).
• Multimodal Alignment: Synchronized visual frames (1 FPS), pre-computed acoustic embeddings (Wav2Vec 2.0), and semantic transcripts (Whisper).
• Statistical Robustness: All 3,000 videos are filtered by a strict exposure threshold ($\tau \ge 3,000$) to ensure reliable performance signals.

📊 Statistics at a Glance

• Total Samples: 3,000 unique professional advertisements.
• Domain: Primarily Beauty & Personal Care (e.g., Anti-aging sets, Pro-Xylane series).
• Duration: Ranges from 6s to 46s (Mean: 21s).
• Data Distribution:
  • ROI: Near-normal distribution (peak at 1.0–2.0).
  • CVR: Long-tail distribution (peak at 0.28%–0.55%).

📂 Data Structure & Format

To ensure privacy and copyright compliance, we provide high-level features and processed frames:

• visual_frames/: 540x960 JPEG images at 1 FPS.
• acoustic_embeddings/: 768-dim Wav2Vec 2.0 features (.npy).
• textual_transcripts/: Whisper-transcribed Chinese text + Metadata (.json).
• labels/: ROI, CVR, and second-by-second iCTR values.
• splits/: Standard Train/Val/Test split indices.

⚖️ License

This dataset is licensed under CC BY-NC 4.0. Non-commercial research use only.

---

🇨🇳 中文说明

🌟 核心特点

AdsTrace 是一个专为短视频营销中的时序广告归因设计的大规模多模态基准数据集。它填补了高频多模态内容与真实商业结果之间的数据空白。

• 细粒度时序标签：提供秒级的瞬时点击率 (iCTR) 轨迹，超越了传统的“黑盒”全局指标。
• 真实业务指标：包含来自两年真实电商投放数据的 ROI (投资回报率) 和 CVR (转化率) 指标。
• 多模态对齐：同步的视觉帧 (1 FPS)、基于 Wav2Vec 的音频嵌入以及基于 Whisper 转录的语义文本。
• 统计鲁棒性：包含 3,000 个高质量专业广告视频，通过严格的曝光阈值 ($\tau \ge 3,000$) 过滤以消除随机噪声。

📊 数据统计

指标	数值
视频总数	3,000
平均时长	21 秒
采样频率	1 Hz (秒级)
全局标签	Pay ROI, CVR
主要品类	美妆护肤 (套装、面霜、面膜)

📂 数据结构

为了遵守版权和隐私政策，我们提供处理后的多模态特征：

• visual_frames/: 1 FPS 下的 540x960 JPEG 图像。
• acoustic_features/: 预计算的 Wav2Vec 2.0 嵌入向量 (.npy)。
• textual_transcripts/: Whisper 转录文本及层级营销元数据。
• labels/: 全局 ROI/CVR 及秒级 iCTR 轨迹数据。

⚖️ 许可协议

本数据集采用 CC BY-NC 4.0 协议。仅限非商业性研究使用。

---