search

harvardnlp / seq2seq-attn

Sequence-to-sequence model with LSTM encoder/decoders and attention
http://nlp.seas.harvard.edu/code
MIT License
1.26k stars 278 forks source link

Character vectors #50

Closed alashkari closed 8 years ago

alashkari commented 8 years ago

Hello, I was wondering what model or method is used in this code in order to obtain the character vectors? Are these character vectors available online? (like word2vec or GloVe for word vectors) Thank you.

yoonkim commented 8 years ago

the character model is from https://arxiv.org/abs/1508.06615 roughly, cnn + highway over character embeddings

alashkari commented 7 years ago

I meant that for each character, we have a d-dimensional vector representation; and since d<|C|, one-hot representation is not used here. How are we assigning these vectors to characters? Are they random values?

Thank you.

yoonkim commented 7 years ago

They are randomly initialized and then trained along with the rest of the model.