search

dalab / deep-ed

Source code for the EMNLP'17 paper "Deep Joint Entity Disambiguation with Local Neural Attention", https://arxiv.org/abs/1704.04920
Apache License 2.0
223 stars 50 forks source link

entity emb question #29

Open Harryjun opened 3 years ago

Harryjun commented 3 years ago

then i fond some question! 1.I think this model rely on entity vector more! 2.[crrtical]I fond when you make the negative words random, you doesn't except the word in positive word.that may be some words in positive and negative

octavian-ganea commented 3 years ago

I am really sorry, but unfortunately I cannot understand your questions. Please rephrase them in a more clear language. Hope it helps.