README.md

TransE

Introduction

TransE is a popular method to model relationships by interpreting them as translations operating on the low-dimensional embeddings of the entities. In knowledge graphs, a tuple is in the form of (head, relation, tail). Negative tuples are sampled by corrupting heads or tails in a positive training example.

Key points to build a TransE

• Get training samples using sample functions.
• Encode entity and relation and use triplet loss to train.

How to run

1. Prepare Data

   cd ../../data/
   python fb15k_237.py

2. Train

   python train.py

3. Evaluate

   python eval.py

Datasets and performance

FB15k-237 is a leakage-free version of FB15k, which is usually adopted as one of the benchmark datasets in literature.

Dataset	hit@10
FB15k-237	0.41

Reference paper

Translating Embeddings for Modeling Multi Relational Data