Emory Team Advances Toward Alexa Prize

A team of five Emory students and one Emory professor, known as “Iris,” beat out more than 195 applications from 15 countries to become one of eight finalists in the 2018 Amazon Alexa Prize competition. The team ranked fifth in the 2017 Alexa Prize competition.

Team Iris has received a $250,000 research grant, Alexa-enabled devices, free Amazon Web Services (AWS) and data and could win up to $500,000 and a $1 million research grant.

Amazon tasked the teams with creating a “socialbot” capable of engaging in 20-minute coherent conversations with humans about a range of current events and popular societal topics. Although the socialbot technology will not be integrated into Amazon’s line of Alexa products in a short time, some of the algorithms may be implemented when they’re ready for commercial use after the competition, according to Iris team leader and computer science Ph.D. student Zihao Wang (21G).

Iris’ socialbot will be capable of fluent and coherent conversation, incorporating real-time search and the ability to offer informed advice and news information, according to Wang.

Amazon selected the eight finalists based on their socialbot’s potential contribution to the field of artificial intelligence (AI); the technicality of their approach; the creativity in designing their socialbot; and the team’s ability to execute their proposed plan, according to the competition’s guidelines.

Team Iris, named after the computer process of informal retrieval and information search, is composed of Wang, Sergey Volokhin (22G), Harshita Sahijwani (22G), Ali Ahmadvand (21G) and Mingyang Sun (19C). Volokhin, Sahijwani and Ahmadvand are all computer science Ph.D. students. A recent graduate of the College, Jason Choi (17C), could not be considered an official member due to the competition’s policy of only accepting full-time students, although he served as a team member in an unofficial capacity.

The variety of experiences and diversity that the graduate and undergraduate members bring to the table is one of the team’s distinctive strengths, Wang said.

“Only a few of us are from computer science [before coming to Emory],” Wang said. “This project needs people from many different interests to make it work.”

A panel of Amazon employees and Amazon Alexa customers will evaluate the socialbots during a customer feedback period this May, according to the competition website. Alexa customers will select two socialbots, and the panel will select at least one team to advance to the finals. After a second round of customer feedback between August and October, three teams will remain eligible to compete for the first-, second- and third-place prizes in August 2018. Amazon will award $500,000 to the winning team members to share and a $1 million research grant to the winning team in November.

The team’s socialbot could play a significant role in advancing the development of conversational bots, according to the team members.

“[The socialbot] is much more usable information [as opposed to an internet search on a regular computer],” Sahijwani said. “You could just omit the need to go to a computer.”

The team hopes that their prototype will not only entertain and socialize with users, but also provide them with informational content on general knowledge and topics such as news, sports, entertainment and science.

“The challenge … is how to provide useful, valuable information while still keeping the conversation somehow entertaining,” faculty adviser and Associate Professor Eugene Agichtein said.

Team Iris aims to reach demographics that could benefit from easily accessible technological platforms. Wang said using socialbots like Iris can help individuals with mental illnesses like depression who find it difficult to confide in other people. Another possible use for the bot is for people who live in countries with limited access to the technology, according to the team members.

The team members said they are striving to make a meaningful impact in the realm of AI for users all over the world.

“It proves that we have the capability,” Wang said. “As students, we have the chance to deploy a system in theory to the real world. It feels like we’re making a difference.”

Varun Gupta contributed reporting.

Correction (2/21/18 at 8:57 p.m.): The article initially said that the team beat out more than 200 universities to become a finalist, but they actually beat out 195 applications. A sentence was clarified to say that the socialbot technology will not be integrated into Amazon’s line of Alexa products, but some of the algorithms may be implemented when they’re ready for commercial use. A quote about the students’ backgrounds in computer science was clarified to reflect that they did not have much computer science background prior to coming to Emory.

Correction (2/23/18 at 10:51 a.m.): The article incorrectly stated that there are 15 finalists. There are actually eight finalists.

Correction (2/24/18 at 10:36 a.m.): The article incorrectly stated that the team ranked fifth in the 2018 Amazon Alexa Prize competition. The team is a finalist in the 2018 Amazon Alexa Prize competition, and placed fifth in the 2017 Amazon Alexa Prize competition. 

0 comments