Text-mining, radically democratized

There is plenty of text information in the world. Most of it is unstructured. It is not organized in databases, neatly labeled or tagged. There are no cues for computers to understand what all this information is about. Take customer reviews. Ok, if it’s in a section of a website that says “Customer reviews” this already can be used as a label. But what if a valuable customer opinion about a product is a tweet like “Got a new iPhone. Love it”. How an organization can find it in the ocean of other tweets and make use of it? And If businesses ignore this information and don’t work with it, they lose their competitive edge: the ability to hear the voice of their customers and react quickly.

Most people think working with unstructured data requires an army of data scientists and plenty of computing power. But this is no longer so. Recently there appeared a number of exciting companies which make text analytics accessible to anyone: developers, startups, small and medium businesses. No linguistic or machine learning skills required. Real democratization of text-mining.

One of them, a startup called MonkeyLearn launched its beta-version in September 2014. The startup aims to become “the WordPress of text mining”. Being compatible with all major programming languages, it easily integrates with apps and provides analytics around the text data the apps work with. For example, it can classify your Twitter followers by their interests, analyze whether the users love or hate your product (what’s called sentiment analysis in a more scientific fashion) based on their tweets, automate product classification and product recommendation. Truly invaluable for startups or SMEs willing to better understand their customers, make their social media campaigns more engaging and improve customer acquisition metrics.

Another startup Indico has just graduated from the Fall 2014 class of TechStars Boston. Using Indico’s service developers can accomplish a number of machine learning tasks like sentiment analysis, text tagging, language detection and emotion recognition – all with a single line of code! Indico aims to democratize machine learning and make it accessible to everyone – this is the company’s mission stated on their website. This means, among other things, that a new generation of models powered by machine-learning will be able to flourish even despite a shortage of data scientists on the talent market.

Similar to MonkeyLearn and Indico, Alchemy API offers its users advanced text analysis services that enable developers to build smarter apps dealing with unstructured natural language data. As a point of differentiation, it also offers powerful image recognition. Imagine your customer said something about your product on Facebook and attached a photo or video to illustrate her point. Potentially, with Alchemy API, you’ll be able to extract information both from the text and from the image and structure it in a way that will provide new insights about what customers think about your products. No data will be lost, which is pretty amazing.

There are even completely free tools for sentiment analysis like etcML developed by a group of graduate and Ph.D. students at Stanford.

Hopefully, with the help of this new wave of companies and tools we’ll soon be able to make much more sense of the world around us. Because most information about it is still in the form of texts in natural languages. It’s our primary form of generating, communicating and preserving information, and it’s hard to get anywhere without it.