/MultiTurnNeuralChatbot

In the modern world, a lot of textual data is generated on a daily basis. All of this data can be used to train a deep learning model to respond to queries posed. While many such concepts have already been implemented to respond to the current query, the way the networks are trained is not similar to how humans converse in the real world. In this project, we have developed a new network architecture that will ensure that the network answers the current query not only by considering the current query but also by considering certain important aspects of the previous parts of the conversations.

Primary LanguagePython

sopro-chatbot-ws2019-

  • New architecture to train network which remembers previous aspects of the conversation.
  • Architecture based on Sequence to Sequence Neural models.
  • Applied Global attention from Lunong et al to give importance to the parts of the concluded conversation that would help answer the current query.

OG repo: https://repos.lsv.uni-saarland.de/christine_schaefer/sopro-chatbot-ws2019