BriefGPT.xyz
Oct, 2019
对话转换器
Dialogue Transformers
HTML
PDF
Vladimir Vlasov, Johannes E. M. Mosig, Alan Nichol
TL;DR
本文提出了一种基于transformer架构的对话策略,其自我注意机制沿着对话的序列进行,可以自然地选择性地忽略或关注对话历史。我们比较了Transformer Embedding Dialogue(TED)策略与LSTM和REDP的性能差异,后者是专门设计用来克服RNN的局限性。
Abstract
We introduce a
dialogue policy
based on a
transformer architecture
, where the
self-attention mechanism
operates over the sequence of dialo
→