Monday, 30 April 2018

Sofia the chatbot helps locate halal food for Muslim travellers, 24x7

Source: Microsoft. Built upon Microsoft's Azure cloud platform and LUIS, HHWT’s chatbot Sofia can automatically provide quick answers to questions posed by users.
Source: Microsoft. Built upon Microsoft's Azure cloud platform and LUIS, HHWT’s chatbot Sofia can automatically provide quick answers to questions posed by users.

Have Halal, Will Travel (HHWT) was started because founder Mikhail Goh had trouble finding halal food in South Korea five years ago. 

“If there were websites out there that discussed halal food in South Korea, they were either hard to find, or any information that I came across wasn’t of much help,” recalled Goh, who is now based in Singapore. “Many times, you have to make decisions based on limited information as a Muslim traveller.

“Do I take the trip down? It’s a 40 to 45-minute ride, based on a single grainy picture, and with very little information to rely on.”

The content platform helps Muslims travel with confidence in countries where locating halal food can be a challenge. HHWT currently hosts hundreds of articles and guides online, and also has a mobile app and social media channels. It now reaches over 9 million Muslims from Singapore, Malaysia and Indonesia each month. 

As readership and site traffic on HHWT continued to grow, it became clear that the platform needed to grow its backend infrastructure in tandem. Fielding 8,000 users looking for halal food options in Tokyo over platforms like Facebook, e-mail or website comments would be best served by a chatbot, the company behind HHWT decided.

Combining Microsoft’s Azure cloud and its Language Understanding Intelligence Services (LUIS), HHWT’s chatbot now answers most of the questions that travellers ask through the platform every day. Called Sofia - chosen after users voted for the name in a public poll - the chatbot is online 24x7, allowing the HHWT team to significantly increase their engagement level with users daily.

“We created Sofia simply because people have been asking us questions which we found could be answered largely through automation,” said Goh.

At the same time, through the responses and feedback received from users interacting with Sofia, HHWT can now analyse and contextualise all that input to help provide more personalised content for each individual user, thus serving customers’ needs better.

Sofia was set up in a surprisingly short amount of time. Instead of creating a chatbot from scratch, the HHWT team relied on existing tools that Microsoft provides on the cloud to create Sofia, with the help of technology partner PleoData.

The Azure cloud platform comes with an application programming interface (API) for a QnA Maker. The feature creates a FAQ from existing content. This was the first building block the team tapped on to form answers to commonly-asked questions. Microsoft’s LUIS, a machine learning-based service to build natural language into apps, bots, and devices then enabled chatbot designers to draw on datasets already in existence. This allowed the team to match users' questions, which can come in many forms, with the right answers. 

“I think one of the really big benefits of Microsoft Azure is the QnA Maker,” said Rachel Tan, MD, PleoData. “It allows any company to be able to release a working chatbot in less than seven working days, which is quite an amazing feat.”

It helped that HHWT already had an existing FAQ, which was input into the QnA maker. Though that was sufficient to launch a chatbot as-is, the team went one step further to customise the chatbot so that it could offer more intelligent content to users.

Currently, Sofia uses what is known as a decision tree that lets users narrow down the answers they are looking for. In future, this process will be simpler and more intuitive—the system will automatically answer questions with a deeper understanding, without having to seek clarification.
This next step for HHWT involves sinking its teeth more deeply into machine learning. With larger datasets to draw on after running the chatbot over time, HHWT would be able to develop contextual and personalised content.

Today, HHWT works with tourism boards from countries around the world and global brands to reach out to the Muslim community. As the same time, they are also working with partners to provide functionalities such as the ability to complete travel bookings and payment transactions on the website. With the scalability of Azure cloud, as well as LUIS, powering the Sofia chatbot, HHWT is now looking towards expansion to more countries within Southeast Asia and beyond.

Goh reinforced that their focus over the next two years is to build on their understanding of their consumers' needs and wants. “One of the things we are seeking to improve for our chatbot is its search and response – to give us a better understanding of what people are searching for, and from there, better responses and better content,” he said.

“As a Muslim myself, I feel that Muslims shouldn’t be restricted to just kebabs or nasi biryani,” Goh noted. “To really understand different cultures, you have to eat their food, speak their language, and really get lost in their country. Muslims shouldn’t be deprived from that opportunity.” 


Watch the video introducing Sofia the chatbot

Download the Microsoft Business Owners’ Guide to Technology

Hashtag: #HHWT