Facebook

Customer Service: Chatbots on web sites (BOT), Through the general personal assistants (GPA), Messaging applications (MES)

Scope: Full

System: Cloud-based (CLD)S

Summary

  • A team at Facebook AI Research used a dataset containing 5,000 human dialogues to develop chatbots with multiple conversational skills.
  • The three skills are being engaging, knowledgeable, and empathetic.
  • The chatbots can be reached through Facebook Messenger.

Delving Deeper

A team at Facebook AI Research used a dataset containing 5,000 human dialogues to develop chatbots with multiple conversational skills. The researchers explained how they trained the chatbot in their pre-published paper on arXiv. Being engaging, knowledgeable, and empathetic are all desirable general qualities in a conversational agent that the work attempts to emulate. In a statement to the press, Eric Smith, one of the researchers, said, “Our goal in this study was to produce a single model capable of smoothly switching between and blending all three of these kinds of communication.”

Previous work has introduced tasks and datasets that aim to help agents to learn those qualities in isolation and gauge how well they can express them. But rather than being specialized in one single quality, the researchers claim a good open-domain conversational agent should be able to seamlessly blend them all into one cohesive conversational flow.

The researchers indicated that their experiments show that multi-tasking over several tasks that focus on particular capabilities results in better blended conversation performance compared to models trained on a single skill, and that both unified or two-stage approaches perform well if they are constructed to avoid unwanted bias in skill selection or are fine-tuned on a new task.