Virginia Tech HokieBot Participates in Alexa Prize SocialBot Grand Challenge 5 to Develop Advanced Artificial Intelligence Technology That Improves User Conversation

Virginia Tech HokieBot Participates in Alexa Prize SocialBot Grand Challenge 5 to Develop Advanced Artificial Intelligence Technology That Improves User Conversation

A team of four graduate students from the Sanghani Center for Artificial Intelligence and Data Analytics are one of nine international university teams selected to participate in the Alexa Prize SocialBot Grand Challenge 5 sponsored by Services LLC. Each team received a $250,000 prize to develop a skill that can help Alexa converse with users about popular topics and current events for at least 20 minutes while earning a user rating of at least 4. 0/5.0.

All challenge entries were reviewed by a panel of Amazon employees and selected based on criteria of potential scientific contribution to the field, technical merit of the approach, novelty of the idea, and assessment of the team’s ability to execute the plan.

Lifu Huang, an assistant professor in the Department of Computer Science and a professor at the Sanghani Center, serves as an advisor to the Virginia Tech team that will build a HokieBot that integrates functions ranging from daily chatting to responding to both open domain and domain specific. questions to provide a seamless user experience, similar to chatting with real humans.

“To achieve our goal of creating an engaging user experience,” Huang said, “we also need to make sure our HokieBot understands the user’s intentions, reactions, and feelings between dialogs.”

By designing a new method for modeling a dialog system that allows for long-term memory, user adaptation, and more controllable generation, the team will address the challenges associated with applying consistency and the increase in human-machine conversation.

Huang offered the following example to show how conversational artificial intelligence (AI) can be successfully combined with consistent, natural responses that include historical context and relevant knowledge retrieved:

User: Alexa, let’s talk

HokieBot: Hi, it’s Linda. Thanks for talking with me. How are you today?

User: Well, not bad. It’s cold now.

HokieBot: Yeah, it’s October. Winter is coming.

User: It’s quite different from my hometown.

HokieBot: Where are you from?

User: Arizona.

HokieBot: Cool! I have heard that the tree-like cactus can grow over 40 feet tall.

User: Exactly. The deserts surround my city and the cacti live better than the tree.

HokieBot: It must be a fantastic and different experience living there. Do you hike? I know there are several mountains over 10,000 feet in Arizona.

User: Yes, a lot! I love hiking with my friends. Oh, and my dog.

HokieBot: Please take me next time too!

User: No problem.

HokieBot: I can’t wait to be there. However, today is not good for outdoor activities. How about watching movies?

User: Good idea. Do you know “I am Sam?”

HokieBot: Yes, it was released on December 28, 2001, a very touching family drama, and I remember it being very long.

User: How’s the movie’s box office?

HokieBot: I’m not sure, but I know he was nominated for a number of awards including the Young Artist Award in 2002 and a number of nominations including the 2002 Oscars.

User: Awesome. Who is the director?

HokieBot: The director is Jessie Nelson. You know, she has a unique eye for casting actors.

User: Yes, Sean Penn did a great job.

HokieBot: Well, he’s won a lot of awards, including the Best Actor awards at the 55th and 60th Venice International Film Festival and the Golden Globe Award for Best Actor. It is awesome.

User: So cool. I can’t wait to watch it.

HokieBot: Let’s do it!

The four doctorates in computer science. the students working on HokieBot are: Ying Shen, who serves as the team leader, and Minqian Liu, Zhiyang Xu, and Barry Menglong Yao. Each has extensive research experience in natural language generation, information extraction, text comprehension, and deep learning.

“We believe that our HokieBot project could potentially outperform existing systems and provide more comprehensive and reliable insights into conversational AI, its potential and its limitations,” said Huang, who also leads a Virginia Tech Sanghani Center team participating. in the Alexa Prize SimBot Challenge. , which focuses on advancing the development of next-generation virtual assistants that continuously learn and gain the ability to perform common-sense reasoning to help humans perform real-world tasks.

The Alexa Prize SocialBot Grand Challenge 5 started on November 14 and the final results will be announced in August.

#Virginia #Tech #HokieBot #Participates #Alexa #Prize #SocialBot #Grand #Challenge #Develop #Advanced #Artificial #Intelligence #Technology #Improves #User #Conversation

Leave a Comment

Your email address will not be published. Required fields are marked *