Open captify-dieter opened 5 years ago
How can pre-trained entity embeddings be used as initialized embeddings rather than sampling from uniform distribution (as stated in the paper)?
How can pre-trained entity embeddings be used as initialized embeddings rather than sampling from uniform distribution (as stated in the paper)?