Natural Language Processing (NLP) is one of the most interesting areas of Data Science. From analysis of the political arena, to organizing meetings, and forming the bedrock of the dream of strong A.I, training computers to truly understand the nuances of human language is part of the yet unreached zenith of Machine Learning. Ahead of ODSC East where Emily Schumm will run a workshop on the subject, here are five cool applications of NLP.
Of the things that we share as a global community, blindly agreeing to a company’s Terms of Service is one of the strangest. We know we should read them, but the time sink proves to be too much of a burden. Using time saving algorithms to get the key ideas out of a piece of text – whether it be Terms of Service, another legal document, or a plain old news article – is an important area of NLP. Services like IBM’s Alchemy API and Fast Forward Lab’s Brief – among others – are taking on the challenge.
Text To Music
TransPose is a tool built by Hannah Davis and Saif Mohammed to construct musical representations of novels. The process involves sentiment analysis on four distinct chunks of the text (beginning, early middle, last middle, and end) and then the sentiment associated with each is used to compose music that matches the tone of the book. It is striking how well it works, though results are not always perfect. In the near future, a similar algorithm could generate movie soundtracks from the script or adapted text.
This is far from mind-blowing in a world where Google Translate is so well-known, but it is also not a big leap to say that most people use it without thinking of the underlying process. This project from Bugra Akyildiz shows the process of building a model to parse eight different languages using data from Wikipedia and scikit-learn. To present this as a look behind the curtain would be overblown, but it’s great to see how common tools can perform these kind of tasks without the power of an entire company at hand.
Startup Textio made a big splash recently with a presentation of their business model to improve job posts. Using Natural Language Processing, a listing is assigned a score which reflects how well-written it is. The analysis gives recruiters feedback on the sections which are good or need to be improved by changing the language used. Though the product is aimed at recruiters, it is sure to be a boon to job seekers as well. The job search process is hard enough without having to read confusing and cliched posts.
Chatbots seem to be the one of the latest crazes in technology. Where bots were mostly ubiquitous on Slack and Twitter, companies like Facebook and Microsoft are pumping resources in developing more intelligent bots. Then there are the growing number of startups centered around providing AI assistants for tasks like organizing meetings and ordering food. (At this point in time, the prevailing notion is that most of these nascent companies involve much more human than machine in the human-machine interface.) This is not a surprise given that tools like deep learning have opened the door to more nuanced products in the field. Chatbots are a necessary step on the path to the goal of strong A.I, and as they grow in complexity, so does this goal come ever closer.
These are just a few of the awesome ways NLP is being used right now. There are sure to be a couple that should have made the list that are absent. Don’t worry, though. They will definitely come across your radar soon enough. NLP is here to stay.