Feb 26, 2013 · 2 minutes

A Bethesda, Maryland-based startup called Synapsify is today launching a B2B product that puts a new spin on semantic analysis, using a patented technique to assess credibility and quality based purely on scouring blocks of text. Synapsify has also announced that it has nabbed $600,000 in seed funding, including a strategic investment from ICG Ventures, the investment arm of book-distributing giant Ingram Content Group.

Synapsify CEO and co-founder Stephen Candelmo, who was previously co-founder of government contracts marketplace Fedbid, says the startup’s technology doesn’t need to rely on metadata, nor does it just do a straight sentiment analysis to assess whether or not a piece of written content is negative, positive, or neutral. Instead, it looks for linguistic patterns and structures that indicate how credible that piece of content is, how high quality it is, and which passages are most likely to resonate with readers. Candelmo likens it to noticing transitions between musical chords that stick in the listener's memory.

“At the very heart of language is sound, and at the very heart of sound, like classical music, is math,” says Candelmo. “That helps us curate content to find the most important and most resonating things that are taking place in that content.”

The idea behind Synapsify is that it will help publishers, marketers, researchers, and companies that deal with large volumes of reviews make sense of content overload. For instance, it can distinguish between a useful restaurant review and a useless review by looking at how the writer constructs an argument and backs it up with facts and detail. “It’s able to recognize a pattern of talking where you are supporting yourself,” says Candelmo. “But if you’re going back and forth and switching from topic to topic and not really crafting a good story, the algorithm recognizes it as a rant.”

The technology is similar to that used by the Huffington Post in its recently launched Conversations platform, which seeks to sort comments by relevance using a semantic analysis engine called JuLia, which was first developed by Adaptive Semantics, acquired by HuffPost in 2010. HuffPost has since kept the technology close to its chest, however. Other competitors include Topsy and Lymbix.

To provide an example of how the Synapsify works, Candelmo put it to work on my story from yesterday about The Magazine. On a scale that, in order, includes “Low,” “Average,” “High,” “Strong,” and “Outstanding,” Synapsify decided my story was Strong in quality, Strong in quotability, and High in credibility. The two phrases that it picked out as most important were:

  • Their models are set up so that they need only a small subscriber base to be able to cover their costs and, all going well, make a little extra money on top”; and,
  • “However, many writers – especially magazine writers – want to do more than just get paid for their work.”
My assessment of the ranking? Not bad. I wouldn’t encourage anyone to read only those two passages instead of the entire story, but they’re good enough to give a general sense of what the piece was about.

Synapsify is a recent graduate of Washington DC’s Fort accelerator and was awarded a patent for its technology three months ago. It has a team of five programmers. Candelmo's co-founder is Lawrence Au, who has a background in search and artificial intelligence. Its other investors include Washington DC venture firm Fortify, Middeland Capital, and some angels.