Knowledge can come from anywhere - classrooms, books and articles. But most times, when we are in a rhythm at work, wouldn't it be super useful to just ping an expert?
Plus, not all knowledge-work is equal - there're apps for buying things, or finding tickets to places. Google is great at feeding curiosity, but goal-driven knowledge is a different game. Situations like interview-prep, sales meetings, project introductions, launching startups, professional development and research can be time sensitive. You almost need a friend to help out with wiping out the noise.
Introducing Mya's KnowBot - a knowledge-savvy conversational bot that understands, learns and helps us with our knowledge-work.
What it does
You converse your knowledge needs with KnowBot by text or call (you'll receive the contact info when you register!).
Check out the video https://vimeo.com/163047019
Talk your heart out, or keep it short and simple - KnowBot is savvy enough to understand phrases, links, and even wordy rants to decipher knowledge-focused needs. KnowBot can understand:
- paragraphs describing your project or work needs
- short phrases you can piecemeal, describing individual facets of your need
- links to job-postings, articles, and anything on the web.
- if you're driving, just call Mya's number and leave a detailed message about your knowledge task at hand!
When you're done, KnowBot intelligently creates a personalized "Springboard Kit", just for you. Today, this kit is available as an easy-to-read packet of relevant articles and expert advice from Mya's design corpus, curated in a pretty-to-see and print-ready magazine. Scan all relevant knowledge on your phone, or print it and take it to-go!
How I built it
Three big, but super fun, steps: 1.) Relevance - this is where using Mya's capabilities to find relevant knowledge really helped us boost the basic premise. 2.) Contextual training the bot - KnowBot is powered by a large pool of real-life training data - a highly curated map of knowledge intents from over 150,000 real-person situations in Mya (http://myalearning.com). So what? The bot can distinguish nuances in intent and context like "I'm moving to Pittsburgh" v/s "I'm meeting with people from Pittsburgh". 3.) "Lego" the cognitive - We've chained a sleuth of in-house AI modules (thinkCOCO prediction and machine learning models) and a number of IBM Watson APIs (Dialog, NLC, Speech-to-Text, AlchemyLanguage, AlchemyData, Document Conversion).
IBM Watson APIs used
Dialog: for basic conversation routing (training mode) NLC: for training intents from Mya's 150K+ real-life situations. Speech-to-Text: call AlchemyAPIs - Language & Data Document Conversion
Mya is an IBM Watson Ecosystem Partner
Accomplishments that I'm proud of & what I learned
- Simplicity is key: conversational UX has an inherent challenge to simplify without losing the human element. It was fun and also great learning to make the Bot smart enough to bridge around like a person.
- Think Platform: KnowBot is fetching public domain knowledge, as well as allows experts to 'plug-in' their expertise. The vision is now near where I could follow my dad, my mentor and key brand-experts - and let KnowBot figure out what's relevant, when and in what capacity.
- Chaining Watson APIs: chaining APIs from a cognitive ecosystem was key to a 'wow' reaction.
- "Just KnowBot it!": I'm addicted to the bot!! And so are our beta-testers - with cool "KnowBot" lingo being thrown around and them printing personalized kits (one of them used the kit as classroom reading for their flight!)
What's next for KnowBot
Personalization. Personalization. Personalization: Follow experts, connect friends, import your LinkedIn profile, and more to see relevant knowledge scores (ask KnowBot things like how relevant is this job posting for me? Should I be pursuing this project?) and get kits that help you bridge the gap.
Enterprise Ready: By enabling enterprises to have a way to feed custom corpuses into Mya, the KnowBot can seamlessly become a learning and knowledge platform for enterprise teams!!! We're excited to have signed up a pilot for this use case.
More platforms: Our bot-framework is compatible with text, voice and we're also testing (in beta) it's natural integrations with Slack, Skype and Facebook Messenger! (all are working, just not release ready, yet - because of minor UI nuances and additional UX we can incorporate for each platform).
This is just the beginning for KnowBot.