These are challenging times. As COVID-19 forces us into social isolation and increased distancing, tensions worldwide are rising and many find it tough to cope with the situation. With more people on the prowl for ‘home activities’ online, it is clear that everyone is looking for help. With ‘Ask Phoebe’, we can help the world cope with isolation better, one home at a time.

What it does

'Ask Phoebe’ is an AI-driven chatbot that offers a treasure trove of personalised social-distance friendly activity ideas while helping businesses market their services, keeping them in business. Whether you’re a parent with 3 kids or a teenager looking for something to do at home, ‘Ask Phoebe’ will recommend suitable activities based on your life stage and preferences.

How we built it

The chatbot was built on TAIGER’s Converse, a chatbot-building platform leveraging AI technology. TAIGER’s Converse is developed using a number of technologies such as Java, Python, Javascript, and VueJS. It also uses Natural Language Processing (NLP) and Natural Language Understanding (NLU) powered by TAIGER’s proprietary AI Platform. TAIGER's Hybrid AI approach leverages Machine Learning, Neural Networks and Knowledge Graphs to understand and process information like a human, so as to provide an engaging conversation experience. For Ask Phoebe, it is being hosted on the Microsoft Azure Platform.

Challenges we ran into

Structuring the conversation People belong to varying life stages, and have vastly different activity preferences. The first challenge we faced was on how to best segment these stages and preferences in a way that would not just feel natural, but would also accurately identify people’s needs. This meant brainstorming and coming up with questions that would best allow us to identify these life stages and preferences.

Content Curation After we set a structure in place, gathering relevant yet interesting content on activities was crucial. Based on all the possible choices, we would have to ensure that each recommended activity would suit both a particular life stage and mood. This meant being selective in the type of activities we would let the bot recommend, and being sharp by considering the context in which they would be doing the activity i.e. alone or with a friend.

What we learned

Ultimately, when it comes to creating a humanistic chatbot, user testing is of utmost importance. Having users test and interact with ‘Ask Phoebe’ resulted in feedback that allowed us to refine the flow of the conversation, create content that was inadequate/lacking, and work on future recommendations.

What’s next

To spread the message, we’re kickstarting a social media movement to invite everyone to #makehomesgreatagain. Collaborating with various government agencies and companies, we are encouraging people to share their favourite moments at home on social media and persuade others to do the same.

We will also be adding search and extract connectors within search engines and social domains to enable the bot to conduct search and curation so as to reduce manual intervention from administrators as much as possible.

Ultimately, we hope for ‘Ask Phoebe’ to be an all-in-one tool where users can receive activities, add it into their calendar, and even make direct transactions on.

What we've accomplished so far

  • High interest reflecting the relevance of our platform. At the initial period of launch, we saw close to a thousand users on the bot without paid media exposure.
  • Global to local media exposure We were picked up by local and regional media like CNA 938, Yahoo, AsiaOne, Technode, Business Insider, and Escape.
  • This activation saw pick up from a wide spectrum of society. From experts in the allied-health sector like The Psychology Practice to community groups like Real Mum SG.
  • Benefiting the larger ecosystem. Through our B2B engagement program which works with corporates to adopt 'Ask Phoebe' within their business strategy, we see SMBs like GUSH, a local paint store and MNCs like Publicis Media, a global leader in media technology leveraging this activation as a marketing and staff engagement tool respectively

Built With

+ 23 more
Share this project: