SuperChat: Dialogue Generation by Transfer Learning from Vision to Language using Two-dimensional Word Embedding and Pretrained ImageNet CNN Models


May 16, 2019 Blind Submission readers: everyone
  • Keywords: Chatbot, Dialogue Generation, Two-dimensional Word Embedding, Transfer Learning, CNN Model
  • Abstract: The recent work of Super Characters method using two-dimensional word embedding achieved state-of-the-art results in text classification tasks, showcasing the promise of this new approach. This paper borrows the idea of Super Characters method and two-dimensional embedding, and proposes a method of generating conversational response for open domain dialogues. The experimental results on a public dataset shows that the proposed SuperChat method generates high quality responses. An interactive demo is ready to show at the workshop. And code will be available at github soon.
  • TL;DR: Print the input sentence and current response sentence onto an image and use fine-tuned ImageNet CNN model to predict the next response word.
0 Replies