search

thunlp / OpenKE

An Open-Source Package for Knowledge Embedding (KE)
3.83k stars 985 forks source link

How to split train/valid/test #306

Closed SRL94 closed 3 years ago

SRL94 commented 3 years ago

Hi, how do you split the knowledge graph into train/valid/test? I am confused about how to make all entities appearing in the training set.

RucLee commented 3 years ago

I have the same problem~! please help

LiuYuemei111 commented 3 years ago

The benchmarks are selected from the original published papers. For example, FB15K Translating Embeddings for Modeling Multi-relational Data FB15K237 Observed versus latent features for knowledge base and text inference WN18,WN18RR Convolutional 2D Knowledge Graph Embeddings NELL-995 DeepPath: A Reinforcement Learning Method for Knowledge Graph Reasoning YAGO3-10 YAGO3: A Knowledge Base from Multilingual Wikipedias You can find more details about how to split these benchmarks.

Hi, how do you split the knowledge graph into train/valid/test? I am confused about how to make all entities appearing in the training set.