Processed MIND Large is a preprocessed version of the MIND Large dataset designed for multi-modal news recommendation research. The original MIND dataset contains news articles with rich textual content and user click behaviors. This specific version is formatted for tasks involving graph neural networks and recommender systems.
Use Cases
- Train a graph neural network on user-article interaction edges to predict click-through rates.
- Develop a multimodal recommender that fuses article text features with user behavior sequences.
- Benchmark recommendation algorithms using preprocessed news article embeddings and user history logs.
- Study cold-start recommendation by analyzing article content features for new items.
Strengths
- Based on the MIND Large dataset, which contains millions of news articles and user interactions.
- Preprocessed specifically for multimodal and graph-based recommendation tasks, reducing initial setup effort.
Limitations
- Specific preprocessing steps, data splits, and feature engineering details are not provided.
- The size, row count, and completeness of this processed version are unknown.
Provenance
- Source
- Microsoft News Dataset (MIND).
- Collection Method
- Preprocessed from the original MIND Large dataset collection.
- Time Range
- null
- Freshness
- null
- Geography
- null