ConMask: Open-World Knowledge Graph Completion and Embeddings

bxshi released this repository in 2019 to support the ConMask model for open-world knowledge graph completion. It provides data and TensorFlow code for representation learning and network embedding, specifically targeting entities not seen during training.

Use Cases

• Generating network embeddings for entities using the ConMask model architecture
• Performing knowledge graph completion on unseen entities using text-based descriptions
• Benchmarking representation learning models for relational data in TensorFlow

Strengths

• Released under the MIT license for unrestricted academic and commercial use
• Provides specialized data structures for open-world knowledge graph completion
• Includes a TensorFlow-based implementation for representation learning

Limitations

• Lacks documentation regarding specific row counts or dataset scale
• Temporal staleness with no updates since February 2019
• Absence of sample data or explicit schema definitions in the repository

Provenance

Source

bxshi via GitHub

Freshness

Last updated February 2019