🧠 NLP and TensorFlow: Text Prediction and Generation
I spent a great deal of time in the last couple of days diving more deeply into deep learning. And I LOVE it! The possibilities are only as limited as my own imagination, and therefore, they are truly infinite. Well, they are as infinite as my ability to stay awake. Because I have to pull myself away and make myself sleep after 14 hours straight. That is how intriguing these infinite possibilities are.
I have always loved languages, because our language forms the way we think and see the world around us. It forms our reality. And so the concept of working with machines to digest and manipulate human language is a frontier that I ran toward full-force as soon as I discovered it. The technology is definitely daunting, because of its limitlessness. And if I am being honest, I am my views are greatly affected by my interactions with Siri and Alexa and AI generated articles on the interwebs that make me often wonder, "Seriously?! THIS is how far we have gotten?!" But now that I am the one working with the machine, I LOVE that THIS is how far we have gotten. Because that means I get to be a part of going the rest of the way! I get to be on the frontlines. The final frontier has never been out there somewhere. It is right here in our very hands.
This is only the beginning for me, so this work is quite simple right now. But I can see so many possibilities, so many ways to improve upon and implement this technology to benefit and serve humans as we swim through the seemingly endless oceans of data that our world has become. So far, this technology has been used primarily as a means of measuring up humans for the purpose of marketing to their subconscious. But that is not all it is good for. And I look forward to being a part of proving that!
Below is an excerpt from the work I did today, a TensorFlow model predicting Yelp reviews. I was overjoyed at my almost 95% TESTING (!!!!) accuracy rate! I have my intuition to thank for that!
And this is a little project working on predicting text in the style of Shakespeare. While the accuracy came out the around 87%, there is still A GREAL DEAL of room for improvement. But this was a great start!
Special thanks to Harshit Tyagi and his fantastic data science resources!