Inspiration
Text Generation is a very interesting AI related topic & involves simulating real speech. My goal was to create a generator which could accurately portray celebrities (based on their music).
What it does
It allows users to speak to Weird Al and Alice Cooper through text generation. Read messages & easily swap between the two while chatting.
How I built it
I created a web parser to grab all the data on a celebrity's lyrics, which I split into the words & fed into my custom generator. The generator then detected what words were appearing in order (following the Markov Chain algorithm) in order to become aware of their speech patterns. This allowed it to generate real text based on what it would expect them to say.
Challenges I ran into
The generator ran into issues with not generating correct data which was an interesting issue. Otherwise it was mostly UI based problems.
Accomplishments that I'm proud of
This was my first real React app that I built (for the front-end). I'd also read up quite a bit about text generation & being able to implement something was interesting.
What I learned
An actual implementation for a text generation technique (Markov Chaining). This gives me a module which I can use for a ton of other purposes. Text generation, in general, has always been a really interesting topic for me & I'm glad to have been able to implement it.
What's next for Text Generator Room
I'd like to make some UI improvements (like storing past messages when you leave for another session). The generator works well, but I believe that could also be improved to be more accurate.
Log in or sign up for Devpost to join the conversation.