When Gokhan Tur joined Uber AI, he started building a team to add conversational capabilities to the Uber app and to do research on dialogue systems. I was the first to join his team, and have greatly enjoyed working with him and learning from such an accomplished expert in the field.
Under his supervision, we worked on both conversational applications and research.
On the applications side, we added voice capabilities to the Uber driver app. The safety benefits for drivers were highlighted in a newsroom article and covered by the press:
- https://mashable.com/article/uber-new-safety-features-voice-addresses/
- https://www.therichest.com/gadgets-and-tech/uber-voice-control-addition/
- https://www.rideshareconnection.com/uber-adds-new-safety-measures/
- https://www.makeuseof.com/tag/uber-adds-new-safety-features/
- https://www.businesstraveller.com/business-travel/2018/09/14/uber-adds-safety-features-for-passengers-and-drivers/
On the research side, we published substantial work in dialogue and NLP. We also developed Plato, a research dialogue systems framework, to accelerate our own work with an extensible and powerful tool.
Some highlights of our work together are discussed in the blogpost summarizing Uber AI's advancements in 2019.
Here is a list of papers the team published on these topics:
- Incorporating the Structure of the Belief State in End-to-end Task-oriented Dialogue Systems, Conversational AI Workshop @ NeurIPS 2018
- Parallax: Visualizing and Understanding the Semantics of Embedding Spaces via Algebraic Formulae, ACL 2019
- Flexibly-structured Model for Task-oriented Dialogues, SIGDIAL 2019
- Collaborative Multi-agent Dialogue Model Training via Reinforcement Learning, SIGDIAL 2019
- Modeling Multi-action Policy for Task-oriented Dialogues, EMNLP 2019
- Plug And Play Language Models: A Simple Approach to Controlled Text Generation, ICLR 2020
- Plato Dialogue System: A Flexible Conversational AI Research Platform, arXiv 2020