search

thunlp / SDLM-pytorch

Code accompanying EMNLP 2018 paper Language Modeling with Sparse Product of Sememe Experts
25 stars 5 forks source link

训练句子向量 #9

Closed xudekuan closed 4 years ago

xudekuan commented 4 years ago

请问，能否用您的模型训练句子向量？谢谢！

wmyw96 commented 4 years ago

Thank you for the question.

However, it should be noted that our proposed module is a language model decoder, which takes context vector as input and give prediction over tokens (see Fig 2 as an illustration). Extending this to sequence embedding might be possible but is far beyond the scope of our paper.