IrisBot : An Open-Domain Conversational Bot for Personalized Information Access


Cited 0|Views5
No score
We describe IrisBot, a conversational agent that aims to help a customer be informed about the world around them, while being entertained and engaged. Our bot attempts to incorporate real-time search, informed advice, and latest news recommendation into a coherent conversation. IrisBot can already track information on the latest topics and opinions from News, Sports, and Entertainment and some specialized domains. The key technical innovations of IrisBot are novel algorithms for contextualized classification of the topic and intent of the user’s utterances, modular ranking of potential responses, and personalized topic suggestions. Our preliminary experimental results based on overall customer experience ratings and A/B testing analysis, focus on understanding the contribution of both algorithmic and surface presentation features. We also suggest promising directions for continued research, primarily focusing on increasing the coverage of topics for in-depth domain understanding, further personalizing the conversation experience, and making the conversation interesting and novel for returning customers. 1 Background and Overview Our goal is to develop a conversational agent that helps the user be informed about the world around them, while being entertained and engaged. Our envisioned social bot, IrisBot, will incorporate realtime search, informed advice, and latest news recommendation into a fluent and coherent conversation. IrisBot aims to discuss and share information on relevant latest topics and opinions in the News, Sports, Entertainment, and general knowledge, by accurately detecting user’s intent, both explicitly stated and implied from the conversation context. Finally, the information will be made more engaging and entertaining by incorporating humor and detecting emotional cues from the user’s utterances. To accomplish this, IrisBot aggregates search and recommendation over a variety of information sources, and employs proactive recommendation strategies, simulated emotion, and incorporate sequence to sequence learning to optimize conversation fluency. At the end of the conversation, our goal is for the users to be informed by in-depth discussions of topics of interest, while entertained through expressions of empathy, humor strategically placed to liven up the conversation.
Translated text
AI Read Science
Must-Reading Tree
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined