search

nju-websoft / JAPE

Cross-Lingual Entity Alignment via Joint Attribute-Preserving Embedding, ISWC 2017
http://ws.nju.edu.cn/jape/
MIT License
98 stars 22 forks source link

Interpretation of output #8

Closed Vika-meow closed 3 years ago

Vika-meow commented 3 years ago

Hi! I tried to use this model and run se_pos.py, attr2vec.py and ent2vec_sparce.py. So, I have output files attr_embeddings, attrs_meta, attrs_vec.npy, ents_embeddings_1, ents_embeddings_2, ents_sim.mtx, ent_vec_1.npy, ent_vec_2.npy, kb1_ents_sim.mtx, kb2_ents_sim.mtx. Can you help me to understand their meaning and explain how do you analyze results?

Best wishes, Vika

sunzequn commented 3 years ago

Hi Vika,

Sorry for my late reply, and thanks for your interest in our work.

"attr_embeddings", "attrs_vec.npy", "ents_embeddings_1" and "ents_embeddings_2" are the output embeddings of attributes or entities.

"ents_sim.mtx" is the pair-wise similarity matrix (calculated based on attribute embeddings) between the source KG's entities and the target entities.

"kb1_ents_sim.mtx" is the pair-wise similarity matrix (calculated based on attribute embeddings) between the source KG's entities.

Vika-meow commented 3 years ago

Sorry for my late reply. Your answer is very helpful, thank you!