Inspiration
I started to read about cognitive commerce some months back and dreamt of a user interface (aka chatbot) that could let me browse shirts from different stores easily by just asking it to show me shirts based on colors, prices and brands.
What it does
It shows shirts based on colors, prices, brands and stores. It tells you the colors and brands available and helps you find shirts using simple language.
How I built it
I use scripts that scrape shirts details and links-to-images from different online stores and load them into mongodb collections, then the chatbot let's me traverse the information in the collections along the 3 shirt-dimensions of colors, prices and brands serving images from online stores and linking to their shopping page.
Challenges I ran into
I wrote lots of lines to convert online stores pages to json files, spent a good number of hours trying to figure out how to correctly get intents with advanced colors like 'spring blue slub' or 'gris dauphin', a common scenario since the chatbot looks into shirts from online stores in USA, France, Germany and Mexico hence lots of different ways to say colors. Building the card responses took me some time too.
Accomplishments that I'm proud of
That it works! And that I can easily get an answer to what's the most expensive shirt at Nordstrom -- It's a python shirt, not the programming code.
What I learned
That Amazon Lex is a powerful tool to capture intent, some Python, AWS Lambda, MongoDB, html scraping and that gray comes in wolf gray, shark gray and dolphin gray.
What's next for ShowMeShirts!
I would like to add women's blouses, dresses and even pants to the bot. Heck, I even dream of kiosks out of malls where you can select shirts or clothing and have them waiting for you in the dressing room...

Log in or sign up for Devpost to join the conversation.