How to deal with words without valid word embedding

[zhaohengz]

Hi,

I noticed that some words in the test set don't have valid word embedding. How to deal with them? Are these categories ignored during evluation?

Thanks!

[yinboc]

If a node has no word-embedding, we just set the vector to be zeros with same dimension. Details could be found in Glove.__getitem__() in materials/glove.py.

They will not be ignored during evaluation, since their predictions depends on the graph context instead of their own word-embedding only.

[zhaohengz]

That clarifys.

Thanks!