We wanted to see if we could implement and create a bot that was actually funny and make the user laugh! Plus, with the pandemic, we wanted to create a unique way for people to disconnect and relax and smile without needing to leave the house.
What it does
We present you with a funny chatbot who has been trained on speech from the fan favorites of television: The Office, Gordon Ramsay, and Friends. Yes, the bot can definitely tell you a joke, but just like all the true comics, its humor flows out effortlessly during a conversation. And don’t forget to ask about the weather! Don’t say I didn’t tell you it’s funny AND helpful!
- Personality: funny with a side of sass
- Hobbies: keeping you company and helping you out
How we built it
- Gordon Ramsay
- Chandler Bing
- The Office cast
- Chuck Norris
- Dad jokes
- React Native
- Amazon Alexa
- Google Cloud
- Adobe Illustrator
SpaCy for POS tagging and linking, word association and NLTK for sentiment.
To determine conversation flow, we utilized Google Cloud’s DialogFlow and designed a conversation flow, creating custom intents and entities. With each user input, the app pings the DialogFlow API to get the correct response. Additionally, we created an Amazon Alexa skill that has the same functionalities as our app, so the user can have the same great experience using Amazon Alexa.
The backend was deployed on Google cloud platform as serverless cloud functions. These functions had a python 3.7 runtime with 128 MB ram. The segregated nature of the endpoints meant that some user config data was stored in a mongodb instance which was looked up upon invocation.
Challenges we ran into
Initially, we created the Alexa skill and saw that Google offered an easy import tool which sounds too good to be true, and it was - that tool no longer works. So, we spent a lot of time reading and understanding many of the Google Cloud Products in order to successfully manually transfer the dialog model. Interestingly enough, you can code in python for AWS Lambda but not for Google’s Cloud Functions, so we needed to find a workaround for the lack of Python compatibility of Google Cloud Functions.
Accomplishments that we're proud of
We're really proud that our product is actually working! Jokes aside, we learned about integrating different parts together from the application to the cloud and the NLP engine.
What we learned
We learnt how to use stuff that makes it work.FunnyBunny has a good sense of humor!
What's next for FunnyBunny
We wanna move it move it