You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Have you considered working on meta embeddings and embedding imputation? I think fse might practically challenge some deep learning architectures, especially when taking knowledge graph embeddings into account.
The text was updated successfully, but these errors were encountered:
Also highly inspiring is the idea of radixAI, which reprojects fasttext embeddings into numberbatch knowledge graph space (solving OOV problem).
My intuition is, with graph embeddings we can tackle bias in NLP better. https://radix.ai/blog/2021/3/a-guide-to-building-document-embeddings-part-1/
But they did not publish them.
Have you considered working on meta embeddings and embedding imputation? I think fse might practically challenge some deep learning architectures, especially when taking knowledge graph embeddings into account.
The text was updated successfully, but these errors were encountered: