Startup companies are gaining momentum in the business sector and Boston-based indico is an example of one such company. indico turns raw text and image data into human insights. The company prides itself on lowering the barriers to the development, integration, and customization of machine learning algorithms by providing developer-friendly tools for building smarter software. Previously, machine learning required a team of PhDs. indico dreams of a world where machine learning is so widespread, seamless, and human-friendly, it no longer seems like technology.
In mid-December 2015, indico set a new industry standard for sentiment analysis accuracy; the indico model returns 93.8 percent accuracy on the IMDB database, even though it wasn’t trained on it. Hence, the current state of the art is around 92.76 percent. Businesses are already using indico’s tool based on this model to measure and better understand public sentiment towards them. According to the official press release:
Results were measured against the IMDB dataset, an academic gold standard. This accuracy increase is largely the result of using a Recurrent Neural Network (RNN) capable of understanding complex language, such as negation and sarcasm, with higher efficacy than traditional methods. indico’s Head of Research, Alec Radford, says that “RNNs are able to work more naturally on sequential data, like sentences and paragraphs. This allows RNNs to be better able to learn about and understand contextual information, which has traditionally been a weak point for technologies like sentiment analysis.” indico’s CEO, Slater Victoroff, says that this increase in accuracy can have a great impact in business. “Traditional lexicon-based sentiment analysis has been shown to be no more accurate than a coin flip; especially when it comes to analyzing the informal language on the Internet. Most companies using sentiment analysis today aren’t seeing signal, just noise. We’ve not only surpassed the current state-of-the-art, but we’ve packaged it in a way that any developer could use. This new API makes understanding your true public position a reality.”
As their systems progress, indico is gaining partnerships. Namely, they have joined forces with InterlinkONE, which is the Boston-based parent company of AwarenessHub, a platform for sales-based social media management, and promote these companies to become users of SentimentHQ. Victoroff says the partnership is a natural fit “We are proud to be partnering with interlinkONE to bring the power of high-quality Sentiment Analysis to bear in AwarenessHub. We want to show the world that quality really does make a difference.”
Essentially, InterlinkONE plans to use SentimentHQ as a way to give AwarenessHub a more accurate and in depth way to gain insight into brand equity. “indico was the logical choice” said John Foley, Jr, President and CEO of interlinkONE, “their work in the field of sentiment analysis yields highly accurate results, which is why we chose to partner with them. The addition of SentimentHQ allows our customers to see the emotion behind online mentions, giving them a sense of how their prospects view them. It’s the ideal addition to our AwarenessHub software.”
Businesses that use AwarenessHub are able to comb social media for mentions of their business and related keywords, and pull everything together in one place. With the addition of SentimentHQ, AwarenessHub becomes a comprehensive tool for not only finding, but understanding brand mentions and related online activity.
Recently Slater Victoroff, the Co-founder & CEO indico, spoke to the Examiner about his experiences working for the organization. Having spent a previous life writing, with the occasional foray into theatre, Slater has found his calling in spreading indico’s vision and getting everyone just as excited about what the future holds. When Slater’s not busy writing emails, he spends his time meditating, MMA fighting, and trying to help out on StackOverflow and Quora:
Meagan Meehan (M.M.): How and when were you inspired to start indico and why was that name chosen?
Slater Victoroff (S.V.): indico started when Alec and I began participating in Kaggle competitions from our dorm room. This ignited a passion for machine learning in both of us. Eventually we had to come up with a name. After weeks of piecing together results from google translate we realized that picking a name wasn’t one of our strong suits. One of our strong suits however, was machine learning. We scraped data on several thousand public tech companies. We created a Markov model weighting letter transitions by market cap. After the algorithm was run, it spit out indico. We like to say that statistically speaking, it’s the most successful possible company name. We didn’t choose the name, we chose the name because it was what the data told us to pick. Our name was generated by one of our very first machine learning models.
M.M.: What exactly is indico? How does it work and what can it be used for?
S.V.: indico is a set of tools that allows developers to use technology from the frontier of machine learning without requiring expertise in the subject. Our primary focus is on text and image analysis. Given a piece of text we can tell you the author’s political alignment, myers-briggs personality type, and anything they might be interested in. We can sift through mounds of unstructured data and find patterns that would take analysts months with just a few lines of code. We can let you train your own machine learning models with 10-100 x less data than you would typically require.
M.M.: You recently set a new industry standard. What was that experience like?
S.V.: Machine learning is extremely competitive. I think this just goes to show that even when we’re up against the billions of dollars that companies like IBM and Google are throwing into this problem, it’s still possible for the little guys to come out ahead.
M.M.: Thus far, what has been the most rewarding part of working with indico?
S.V.: Talking with happy customers. Nothing will get you up in the morning like knowing that you’re making a difference, and when people tell us that our product is wonderful it inspires us to work even harder.
M.M.: Where do you hope the company will be ten years from now?
S.V.: I hope that in ten years machine learning will be boring. I hope that it will be so commonplace and transformative that we cannot remember before it. I hope that fifty years from now students will learn about the machine learning revolution and know that indico played a part in bringing this technology forward.
* * * * *
For more information about indico visit its official website, Twitter, Facebook, YouTube and Instagram.