search

thunlp / KB2E

Knowledge Graph Embeddings including TransE, TransH, TransR and PTransE
MIT License
1.4k stars 449 forks source link

test for pTransE #60

Closed ShellingFord221 closed 6 years ago

ShellingFord221 commented 6 years ago

I want to predict head by pTransE according to relation and tail, I should:

  1. Use transE to get top 500 candidate entities according to relation and tail
  2. Train embeddings of all entities and all relations by Train_TransE_path.cpp
  3. Make triple by candidate-relation-tail (500 in total) and put them into PCRA.py to get path.txt, confidence.txt, test_pra.txt for test (I can't form e1_e2.txt and use it because there is no relation mentioned in it.)
  4. Use embeddings, path.txt, confidence.txt and test_pra.txt for test by Test_TransE_path.cpp to get pTransE scores and rerank candidates. Am I correct?
Mrlyk423 commented 6 years ago

Yes, you are correct.