Skimbox launches, taps analytics for inbox triage

Skimbox launches, taps analytics for inbox triage

Summary: Can neural networks tame the email beast?

TOPICS: Big Data

Be honest.  You dread email.  It's a volley that never ends, and one that's not exactly efficient, either.  Plus, all those Effective Habits of Getting Things Done can help, but they're not panaceas.

What if machine learning could help instead?  What if statistical models about which email is important to you, based not only on things like volume of correspondence, but also observed behavior, could be applied to separate your email wheat from the chaff?

That's the thinking behind Skimbox, a product incubated at Watertown, MA-based SoftArtisans.  Skimbox provides a native iOS client for Gmail- and Exchange Server-based email that intercepts and segregates your "skim" -- email that you'd likely rather skip, or read later. 

Skimbox is built on the premise of ferreting out "signals" from your email content itself, and from your treatment of it.  Here are some factors that are particularly germane:

  • Header content (especially sender, of course)
  • Monograms, bigrams and trigrams (one-, two- and three-word phrases) that are common in important messages and their subject lines
  • Determining which emails are read
  • Determining which emails are categorized

The result?  An inbox with your important mail separated from everything else, the latter being your skimbox and the former being your mainbox:

skimbox_move to skim


Curious about the technology used to build Skimbox?  I was, especially since I happen to know the company weighed its database options very carefully.  Ultimately, the team settled on MongoDB for the database, Ruby for the back end, node.js for Web service tier, and Objective C for the iOS native app (natch!).   Skimbox's classification engine, meanwhile, is written in Python, using NLTK (the Natural Language Toolkit), with neural net functionality implemented using the Enthought Python Distribution.

Will natural language processing and neural nets be enough to tame the email beast?  Time will tell...or maybe you will, as you can download Skimbox from the App Store now.

Disclosure: I've done work in the past for Skimbox's incubator-parent, SoftArtisans.

Topic: Big Data

Andrew Brust

About Andrew Brust

Andrew J. Brust has worked in the software industry for 25 years as a developer, consultant, entrepreneur and CTO, specializing in application development, databases and business intelligence technology.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Interesting tech. But will it be more or less accurate than Google's tabs?

    Interesting tech. But will it be more or less accurate than Google's tabbed email, which separates email into main, social, promotions, updates, and forums?

    I dunno what the algorithm is exactly, but I'm sure there's some sort of machine learning involved there as well.

    The question is, which algorithm will prove to be more accurate?
    • Re: Skimbox's machine learning algorithms

      The primary way Skimbox classifies email is through a neural net, which sniffs out latent and evident important/non-important signals in your email interaction history. We've written a bit about it here, if you're curious: