A common problem that plagues financial institutions is maintaining the accessibility of online / mobile app based banking services. Unfortunately it can be a bit of a design challenge to help the technologically illiterate navigate endless lists, accommodate the elderly or physically disabled, or assist people who may not speak fluent English.
I devised the Tali Chatbot as a solution to this, allowing users to communicate naturally to a friendly interface that handles any questions they have regarding account data, upcoming bills, or budgeting through a single endpoint. Tali lets users communicate naturally through speech or text as if they were speaking to a real person.
What it does
Tali's functionality allows users to handle all information requests like "What is my account balance?" to "Can you keep track of this new bill?", running it through Microsoft Language Understanding and appropriately returns the answer to a user's question or walks a user through the steps in handling financial status updates.
How I built it
Tali is a C# Chatbot built on the Microsoft Bot Framework. I used Capital One's Nessie API to handle financial data using sample customers, and Microsoft Language Understanding (LUIS) to handle gathering user intents and allowing for natural conversations to form without having to worry about triggering predetermined phrases.
Challenges I ran into
I had originally intended the final stage of this project once all (demo) features were implemented to be a full port to Azure and Microsoft Cortana, but there was issues with getting the correct bot endpoint to register. There was also the concern of handling invalid or random commands, as determining the cutoff between where to draw the line between "highest scoring intent" and "this doesn't really make sense" is hard to pinpoint.
Accomplishments that I'm proud of
In a single Hackathon I learned all about Chatbots in a team of one (myself). I was able to devise multiple solutions to handling dynamic conversations that can form between a user and the bot, and at the end of the day I was able to implement most features I wanted with no prior knowledge of Azure's capabilities or how bots function.
What I learned
I learned a ton about Microsoft Azure's suite and the functionality C# chatbots can offer its users.
What's next for Tali
Tali's next steps involve further language training for its financial suite, and I plan for continued development into other areas from smarthome automation to personal reminders and social updates. I also strongly wish to reattempt a port to Microsoft Cortana and give Tali full integration into day to day computing in Microsoft Windows 10+ through voice commands.