arrow flex-icongithublightnings linkedinlock target triangle-icon twitter

Building a Chatbot for Higher Education | DVELP + Google

Ross
October 30, 2018

DVELP partnered with Google in a Chatbot Hackathon organised by UCISA and JISC, alongside Microsoft, Amazon and IBM.

AI chatbots are very much the flavour of the month and this was reflected in the excitement around the day. There was standing room only at the event as representatives from academic institutions across the country look to leverage chatbots to improve their student and support services and increase efficiency.

Despite their popularity, the experience of using chatbots in the last few years has often been frustrating. A narrow scope of pre-programmed questions, clunky answers and infuriating conversations have often led users to feel it would be much easier to ‘speak to a human’.

However, the functionality provided by natural language processing (NLP), and in particular Google’s machine learning API has led to huge improvements. It aggregates millions of conversations meaning the underlying technology is always improving and providing a better experience to users.

The JISC hackathon was a great chance to explore how to implement chatbots to get the right information to users quickly. And when the bot can’t do that, make sure we route users to the right person to help out!

The event had two main sections. The first was a brief to present a chatbot using real data provided by JISC from their IT service desk. The second was to run a ‘how-to’ workshop which would let attendees go back to their universities and create chatbots of their own.

When introducing ourselves and Dialogflow, we wanted to emphasise the speed at which basic chatbots can be created. We used the Knowledge base feature within Dialogflow to upload the JISC data, process it, then integrate with a Google Home speaker to support live voice conversations.

This meant within minutes, in front of a live audience, we’d set up a chatbot that could respond to all customer queries supported in JISC service desk data.

In the ‘how-to’ section, we wanted to show the power of Google Dialogflow, and how easy it is to use. We’ve been working closely with the University of Portsmouth to build a chatbot for their open days. This chatbot would have the goal of increasing sign-ups and supporting students in the run up to open days. It was crucial that the solution is truly omni-channel, so we used our own platform; Airline. This lets Dialogflow support online chat, SMS, Messenger and Voice with the same intent model. It can even handle routing to a service agent if the bot doesn’t have the relevant information.

After the demo, we took the workshop attendees through the process of building it. We ran a real 'how-to' where the audience created their own agents and built their first intents to started training their bots. The first intent we saw working… “Open the pod bay doors, HAL.”...

If you’re interested in knowing more about making your own chatbots, Dialogflow or the Universities chatbot product then get in touch!

Articles

By using this website you agree to our cookie policy
x