search

WenZhihao666 / G2P2

Augmenting Low-Resource Text Classification with Graph-Grounded Pre-training and Prompting. SIGIR-2023
MIT License
21 stars 2 forks source link

About node features #1

Open yuh-yang opened 1 year ago

yuh-yang commented 1 year ago

Hi Authors,

Thanks for your inspiring work. As I'm trying to reproduce, may I ask which pretrained word2vec vectors are used here for generating node features from texts? It is from GoogleNews or FastText? Or may I have link for relevant information?

Thanks!

HelloWorld4747 commented 5 months ago

Hello, Have you solved this problem? I have the same issue.

Thanks in advance!

yuh-yang commented 5 months ago

I chose to use SBERT to reconstruct the feats.

WenZhihao666 commented 5 months ago

Hi Authors,

Thanks for your inspiring work. As I'm trying to reproduce, may I ask which pretrained word2vec vectors are used here for generating node features from texts? It is from GoogleNews or FastText? Or may I have link for relevant information?

Thanks!

Not these public pre-trained word2vec vectors. I trained these word2vec vectors from scratch by myself, by using the Gensim word2vec package.

WenZhihao666 commented 5 months ago

I chose to use SBERT to reconstruct the feats.

It's ok to use other ways to get the node features. But kindly note that you need to pre-train the whole G2P2 model again.