In one of my projects to obtain more email signups for my role at ODSC, I create a downloadable PDF, the “ODSC Guide to Natural Language Processing,” which contained highlighted blogs and recorded ODSC talks related to machine learning, among other pieces of content. At the end, I wrote a letter from the editor with my thoughts on machine learning.
Graphic design was done by Paxtyn Merten and Ava Burcham.
Content compilation and descriptions by me.
Maybe I’m biased, but I think language is pretty cool. But hey, I’m just a writer.
Really though, I find NLP to be one of the most important developments under the data science and AI umbrella. Humans speak with words, not numbers, and it’s important that we do all we can to teach our machines to understand words and expression.
NLP was probably my first foray into data science. When I was earning my master’s degree, I had to take a course on social data analysis, and in the world of media studies, that meant a lot of Twitter data. Considering my background has always been in writing, learning R wasn’t the easiest – but I found the results fascinating.
In said course, my team and I examined (an at the time trending piece of) research on the consumption of red meat and cancer. Using sentiment analysis, we wanted to see if the public perception of the term “red meat” had any notable change before and after the World Health Organization (WHO) report on red meat emerged.
As you can imagine, not only did more people discuss “red meat” on Twitter (in reference to food), but also the tone of the discussion became more negative. The results didn’t really surprise us, but the process itself was exciting.
Call me crazy, but I loved manual labeling. To teach our machine what to look for, we hand-labeled 200 tweets each, flagging for spam and other non-related content. We performed this research in the summer of 2016, so there were plenty of posts related to “red meat” in the context of politics. It felt oddly therapeutic to clean out all of the unrelated Twitter posts and just leave the relevant ones.
I thank my time at Boston University for making me interested in data science. Since joining ODSC, I’ve only grown to appreciate it more. It’s not just red meat and hand-coding anymore.
When I first learned about NLP, I thought it was just a tool for social media, but I was clearly quite wrong. One of my favorite applications of NLP is definitely in the healthcare setting; being able to develop systems that can diagnose diseases, predict potential risks, and even help with treatment is quite literally a life-saver.
NLP – and of course other AI techniques – will likely never replace humans completely. Though humans must learn to use AI to the best of our abilities. AI is becoming a powerful tool, and with 500 million+ tweets sent daily, we could use all the help we can with interpreting the vast amount of words that we write every day.