Conversational AI at Smart India Hackathon 2020
Over the last few months, MindMeld was thrilled to be the Conversational AI partner for the Smart India Hackathon (SIH) 2020. SIH 2020 is a nationwide initiative by the Government of India to foster technological innovation and exploration among engineering students across the country. The hackathon included a hardware and software edition, with conversational AI featured as a prominent software category.
This year, the focus of the conversational AI category was to promote tourism in India using conversational interfaces to improve the tourism industry’s quality and effectiveness. Participating teams used the MindMeld Conversational AI Platform to develop an intelligent chatbot to assist tourists with various travel needs like trip planning, itinerary booking, activity and sight recommendations, and general logistics management. The participants chose from several scenarios, including adventure tourism, ecotourism, wellness tourism, and cruise tourism.
In the first round, held in early January 2020, student proposals were submitted and a team of MindMeld evaluators carefully examined and scored them according to originality, completeness, and relevance. The top five moved on to the final round, held August 1st through August 3rd, and the teams spent three days building and demonstrating their chatbots.
During the 3-day grand finale, a group of Cisco technology leaders from the US, Canada, and India, including members of the MindMeld team, provided support and mentorship to the students as they developed and refined their solutions. A separate group of judges, also from Cisco, evaluated the final demo presentations and provided feedback.
The evolution of each solution from a proposed concept into a mature solution in just three days was very impressive. Each team implemented a unique chatbot using the MindMeld platform and integrated it with a major chat application, such as WhatsApp or Webex Teams. The quality and originality of the demos were high across the board.
The first prize was awarded to the team from Fr. Conceicao Rodrigues College of Engineering in Mumbai, who implemented a novel conversational agent for cruise management utilizing a variety of advanced MindMeld features such as dialogue flows and automatic slot-filling. The other finalists were teams from Indian Institute of Information Technology, Design and Manufacturing, Jabalpur, Indian Institute of Information Technology, Ranchi, Bajaj Institute of Technology, and Shri Sant Gajanan Maharaj College Of Engineering.
Below is an interview with one of the finalists, sharing their experience participating in the Smart India Hackathon and using the MindMeld Conversational AI Platform to power their award-winning solution.
Tell us a bit about you and your team.
My name is Apoorva Singh, and I am the team lead for CodingDovahhe from IIIT Jabalpur, Madhya Pradesh. Our team was selected for the Grand Finale of Smart India Hackathon 2020 under the Atithi-Devo-Bhava problem statement presented by Cisco DevNet. I would like to thank my team members Himanshu Pandey, Rohan Saxena, Siddharth Verma, Srijan Singh, and Tejendra Singh for their sincere efforts in bringing our idea to life.
Can you tell us about your solution?
In this problem, we had to develop a chatbot that would act as a virtual travel agent using the MindMeld Conversational AI Platform. We started our implementation by brainstorming approaches and deciding on the best among all of them. We also wanted to provide the best UX for the user.
We chose WhatsApp as our preferred chatbot deployment because it provides location sharing, which we wanted in our chatbot. We intended to use this to make the user experience hassle-free by sharing their location without typing it. Our chatbot focused on adventure tourism spots classified by various parameters like season or activities.
The chatbot provided suggestions based on the user’s preferences and location. Firebase was used to save the user’s data, and different users were distinguished based on the phone numbers they used for WhatsApp.
Here’s a short demo showcasing just a bit of what our chatbot offers:
How was your experience working with MindMeld?
Our team loved the versatility of MindMeld as it met our requirements well. The chatbot was powered by MindMeld and integrated with WhatsApp through Twilio’s messaging service. The main server’s architecture was based on Flask micro-framework, which was included in the MindMeld package. Each conversation was personalized to the user based on their stored preferences in the bot.
The ML pipeline was very simple and easy to understand. We even tweaked the hyperparameters for domain and intent classifiers and experimented to optimize accuracy. Overall, the MindMeld platform relieved us from much of the boiler plate involving the NLP classifiers. Building the dialogue manager was trickier. We could implement most of our conversational flows without much hassle but some of our more complicated flows required us to dig into the platform code to understand the inner workings better and tap into more advanced functionality. But overall, the documentation was great — it had comprehensive information about all the features and functions separated out into different sections.
Can you share a few lessons that you learned from SIH 2020?
It was a great experience, and we learned a lot. Participating not only polished our technical skills but also honed our management and team-building skills to a great level. In the end, the journey matters more than the destination. We will cherish this experience and learning for a lifetime.
The MindMeld team congratulates all the SIH 2020 finalists who participated under the conversational AI category and developed innovative, intelligent solutions that will go a long way towards promoting tourism in India.
If you are interested in using MindMeld to build your own chatbot, visit our website to learn more. We welcome every active contribution to our platform. Check us out on GitHub, and send us any questions or suggestions at mindmeld@cisco.com.
