I heard  about Glide Intelligence, an advanced social media analytics platform produced by  Glide Technologies (a Vocus competitor that created a platform for online media analysis and publishing) from Sam Phillips, CEO of Glide Technologies, who showed me a glimpse of  GI is summer and then a much fuller example a few weeks ago while I was visiting Boston for Monitoring Social Media 2010 Boston.

In fact, at the very moment  Sam and Keith showed me the advanced features of Glide Intelligence to me, I got an email from my publisher, McGraw Hill, telling me my book deal had been approved and I took that as an omen that, indeed, I was not only looking at something very new and important, but also that I was exactly in the spot I needed to be in at that moment.

The in-depth tour was given to me in person by Sam Philips and Keith Woods-Holder, Glide Technologies chief software architect.  I’m familiar with Glide Technologies through my speaking engagements in London over the last year and I will be speaking at an event for them later next month while visiting London once again.

GlideIntelligence^TM comprises an online media monitoring and analysis tool you can access through a web browser. The system  incorporates advanced sentiment and phrase tracking to deliver fast and accurate quantitative and qualitative metrics either as charts and text summaries, or as predefined reports.

User controls allow clients to customize and modify system behavior to obtain a personalized dashboard reflecting their own priorities and emphasis for reporting and I find this feature fairly unusual for this type of software and haven’t seen such a feature in any of the other platforms I have looked at or use (though I have heard of this feature of auto-learning in specific applications elsewhere, just not in Social Listening or anything related to Public Relations).

Automated alerts and distribution features also allow users to receive custom lists of key articles and sentiment on a daily basis and this is some of what Glide Intelligence reports on.

• Topic (e.g. environment, results, announcement, customer etc)
• Source type (News, feature , review, analysis etc)
• Spokespeople (named people in stories)
• AVE (calculated value articles if they were paid for as advertising)

(and if needed refine) analysis components to ensure that management priorities are fully reflected in results and reports.

GlideIntelligence^TM has access to over 100,000 online news sources and can incorporate customized feeds of requested materials, including those from ‘pay wall’ and copyright protected sites (with agreement). The social media components of the platform can access all public content and uses customized API interfaces developed in cooperation with the major network providers including Twitter and Facebook.

The core of the analysis is a complete linguistic and semantic analyzer process developed by Glide. The key features are:

• Using an extension of Viterbi logic, Glide’s analyzer does not rely on dictionaries and definitions to work. Instead it analyzes the content for structure and , crucially, context to derive data points for sentiment, tone, relevance and association.

Above is a snapshot of the main dashboard of Glide Intelligence.  Users are initially presented with a home page, at launch it will be pre-configured with a selection of news monitoring as default metrics settings based on the  brief. The Home Page is a summary of key stories, latest news and charts and is designed to give each user a quick overview of results any time they access the system.   As you use the system it quietly reconfigures itself to your preferences mirroring back to the user the views of the data they prefer and making them defaults.

Glide Intelligence has some innovate features I haven’t seen anywhere else such as multi viewpoint sentiment and textual analysis  – here’s an example of what I’m talking about (below):

Glide offers three viewpoints as defaults – this is unusual as you can predefine entities such as officers of a company your monitoring and show how a news story or event is positive or negative to the entities you define and much of that learning comes from machine learning and artificial intelligence – it’s not actually keyword based (but that doesn’t mean keywords can’t be used to get the system started).

• Industry (general language rules apply always)
• Client (all relevant scores including those for competitors are modified to show how sentiment looks for from the client point of view)
• Competitor (which allows users to ‘view’ how the sample looks from a competitor point of view.

This chart above shows how a user can define entities down to the company officer level and the semantic references can be positive and negative relative to the entities you have defined, and this is a part of Glide Intelligence’s “secret sauce”.  It makes total sense  something positive to me might be negative to you and vice versa, yet no other platform I’ve seen has actually come out and find a way to process information that way till now.  Consider how much time such a breakdown of information saves the analyst, saves the company that uses it.

Yet another feature of Glide Intelligence was a report that breaks down the content (equivalent to the  River of News and Topic Clouds in Radian6) in to meaningful multi word phrases that have conceptual meaning, something Radian6 is incapable of doing.

Most platforms can not really breakdown online mentions and categorize that information in a meaningful way, this work is usually done by people but Glide Intelligence can handle it (see above).  Again, this feature can be a big time saver.

Information that Glide Intelligence reports on can be broken down further

• Messages (appearance of key messages in articles; note the system support both
  keyword and key phrase, the latter allowing a more accurate match to sentiment
  based measurement)
• Topic (e.g. environment, results, announcement, customer etc)
• Source type (News, feature , review, analysis etc)
• Spokespeople (named people in stories)
• AVE (calculated value articles if they were paid for as advertising)

When I spoke with Sam and Keith a few weeks ago they asked me what I thought of Glide Intelligence and it struck me how much time and manual, mind numbing work Glide Intelligence saves a firm that uses it and how that may be the most beneficial thing about it.

In my latest incarnation of how I apply analytics, having spent the last 15 months working with Public Relations firms that are learning how to integrate Social Listening and Social Media into their offerings, I’ve noted how labor intensive and manual all of this really data pulling and reporting really is (too often clients  have no idea of just how much work is involved in even attempting to do this kind of work the right way and there is a deep cultural gap between traditional public relations and advertising and real social media and web analytics to the point the “business” typically doesn’t understand what they are buying, what they are asking for, how long it should take and how much it should cost).

Now, it can be argued that anything cutting down on the manual drudge work  is worth considering simply based on the utility of saving your mind for the things that machine learning can’t do which is create the high level synthesis that only humans can, but humans that understand the data they are looking at.  In fact, if we spend too much time (more than 30%-40% of our time manually collecting and assembling data) we will end up being too burnt out to analyze the information or come up with anything truly actionable or that interesting and that is usually the case.  Glide Intelligence drastically cuts down on the time you have spend gathering the data and assembling it and what Keith showed me in Boston blew me away (see the chart below).

A conversation can be tracked over time ( here we are looking at how the subject of conversations threads changes over time  -  the individual points are summarised by the analyser to show ‘topic groups’ and the rate of followers/change etc provides the trigger point for each step in the conversation).

The  visualization above Glide Intelligence creates on the fly with advanced AI data, should a person had to construct this view it may have taken between a half a day to a full day of work, all things being equal, and is of immense value in certain types of reporting.   Doing it the old way an analyst would be brain dead and exhausted by the end of that day and  would not have anything left in them to summarize what this view of the data means -now they can because that view, the view above, is created in a micro second by Glide Intelligence.   I know from first hand experience just how hard it is to pull this information and chart it in a meaningful way; also the act of gathering data is entirely subjective, meaning I may pull different information than another analyst doing the same thing with the same data, Glide Intelligence takes care of that problem by ensuring no one has to create this chart because the platform does it for us.

Another problem that Glide Intelligence attempts to solve is tracking the true impact of a media placement (this is extremely important to PR, for example, that often produce media placements for clients and are then asked to explain, by the client, what that media placement did for client).  Tracking the impact of media placements has been a real problem in that just about no one has any good software that does it nor are any of the Social Monitoring and Listening Platforms able to handle this requirement well.  Firms have tried to use Sysomos and Radian6 to do this kind of analysis of media but those systems are not designed to to provide the detailed reports that is required.  However, Glide Technologies decided to try to solve the media placement problem through Glide Intelligence (see chart below):

Because Glide Intelligence is integrated with Glide’s other platform components the results can be dynamically linked with events or activities for a client and included with other metrics from dissimilar media types.  For example, a user can show the relationship between a newspaper articles and the social media debate it inspired (chart above).

I expect Glide Intelligence will be coming onto the market shortly and I think we need to start moving away from the manual culling of data (because PR can’t do that very well, anyway) and into advance machine learning.  Up to now, machine learning programs weren’t that good for this kind of work, but that view is clearly changing as we see products such as Glide Intelligence becoming available.

Now, I haven’t used Glide Intelligence or any of the other Glide software so I can’t yet tell you what it’s like to be at the cockpit of this platform, but I can tell you that if your a Public Relations or integrated communications firm Glide Intelligence is a platform you’ll probably want to and should be using in-house.

That’s all the information I have on Glide Intelligence, if you need more contact Glide Technologies in London (they now have a NYC office as well).

Monitoring Social Media San Francisco just took place yesterday, October 21st, and Monitoring Social Media Bootcamp is happening today.  Judging from the twitter stream yesterday’s event in San Francisco was great (my guess is this link only will get the content I am referring to for a few days, then will age out of Twitter and won’t be easily accessible).   I particularly regret missing Gary Angel’s presentation on Social Media Dashboards and I knew when I referred he would make a fantastic impression (which he did); hopefully it’s videotaped and if it is, I’ll post it here.

Today is the Monitoring Social Media Boot Camp San Francisco, I attended and spoke at the first event of this series in London on March 31st, this is the second event (today) in SF and the next one will be in New York City (where I live) in two weeks (November 5th).  I am curious to hear how today goes and will be listening to the Twitter stream, the format has changed from what we did in London and will be more hands on and also half a day, not a full day.

Then Monitoring Social Media New York is coming up on November 4th, where I’ll be on a panel about Social Media ROI with Katie Paine at noon.   I did a presentation in Boston two weeks ago but this time I’ll just be doing a panel (sigh).   Following #msm10 is a bootcamp, just like what is taking place in San Francisco today but it will be even better I think, because we’ll have some experience behind it from today’s event and the hands on approach was actually partly inspired by me as it is called a “bootcamp” right?    The bootcamp should be totally hands on with no questions barred, all vendors presenting should be evaluated on the same queries, same time period and approach.

In fact, if it were entirely up to me, I’d have set the presenters 3 tasks

1. a geo local query (lets see how the platforms handle finding something local)
2. an industry segment analysis (all platforms evaluated must use same query and time period) then let’s compare the results and ask them ad-hoc questions.
3. A comparison of Sentiment Analysis for the same query, also measuring volume, tone, segmentation of content to various social media channels such as twitter and facebook.

Will be interesting to see what comes out of this.

A couple of weeks later I’m in London to present at Monitoring Social Media London on November 22nd where I’ll be doing a full presentation again.     I’ll also be speaking at some local events and taking a short trip to Oslo where I’ll meet some interesting Norwegians about Social Media Monitoring.  More about that, later.

Some of my readers may recall my post on Minority Report and Web Journal Sept 30th to October 2nd 2010 where The RecordedFuture was mentioned as in reference to online employee surveillance monitoring.

Interestingly, the article mentions another company I’ve written about recently called Recorded Future (small world), but I didn’t know they were partly owned by the CIA and Google.

A Cambridge, Mass., company called Recorded Future, which is funded by both Google and the CIA, claims to use its “temporal analytics engine” to predict future events and activities by companies and individual people.

The article makes a prediction about predictions ….

Recorded Future is only one of many new approaches to predictive analytics expected to emerge over the next year or two. The ability to crunch data to predict future outcomes will be used increasingly to estimate traffic jams, public unrest, and stock performance. But it will also be used to predict the behavior of employees.

And Google has figured out how to tell when someone is ready to quit Google.

Google revealed last year, for example, that it is developing a search algorithm that can accurately predict which of its employees are most likely to quit. It’s based on a predictive analysis of things like employee reviews and salary histories. They simply turn the software loose on personnel records, then the system spits out a list of the people who are probably going to resign soon. (I’m imagining the results laser-etched on colored wooden balls.)

I have full access to the platform and here’s a review.

Recorded Future tries to answer a question in terms of “Who, What and Where” and leaves you to figure out “Why” and since the event is in the Future (it monitors future expected events based on current and past trends) the platform tries to predict what is likely to happen in the Future (in a time window you specify).    If it can, Recorded Future will predict likely outcomes, as mentioned above Google and the CIA have invested in this platform which I did not know when I first spoke with Chris Holden at Recorded Future.

For the “What” part of your query you have several segments already set up (currently you can’t make your own segments but Recorded Future can make one for you, if you really it want it bad enough).  You can choose from Capital Markets, M&A, Accounting & Restructuring, Guidance and Ratings (financial), Entertainment Industry, Corporate Growth, Patents,  Persons, Products, Legal, Disasters, Pharmaceuticals, Crime & Violence and Government & Politics.  In each category segment there are several sub-segments that allow for honing in.

Let’s select Entertainment with a sub category of Movie Release.

The next question Recorded Future asks of you is “Who do you want to know about” or “Where do want to know about” and that is a field you just fill in.  For example, if I want to know the upcoming releases and movies of a certain movie star, say, Angelina Jolie, Recorded Future will try to predict what they will be based on what it already knows, which is different than what other types of search engines do (they are based on past and present).

Let’s select Angelina Jolie.  If Recorded Future has a record on Angelina Jolie your set (you can choose a few options from among those it finds.  If you do a search on a person Recorded Future doesn’t know about yet, it may start collecting information on that person and you may not get much back from your query when you first execute it in Recorded Future.

The final  question is When is going to happen, What time period is this going to cover.  When I entered “Next” I could specify next day, next week, next month, next year, and so on.  I selected over the Next Year.

The query didn’t find anything on  Angelina Jolie for the next 12 months (hard to believe) but it does allow us to set up a query to look for those events in the future and email you when it finds something.

Next I asked Recorded Future to find out anything about Legal issues with Tiger Woods that happens anytime in the Future (using the same three drop-down boxes) and now we get something, except it doesn’t look like, at least on the surface, to be anything about Tiger Woods that we expected to find.

I’ll need to check back with Recorded Future to understand how I could choose “Legal”- General Topic, Tiger Woods (the person) – I could have also chose his Foundation, etc, and the Future was the next months, but it looks like what I got was anything but Tiger Woods – you tell me.

I created a network graph and it has very nice visualizations you can hone in on and expand but I still find we need to understand the language of platform like this to effectively use it (probably trial and error on my part as I did not get what I expected of the bat – but that doesn’t mean I can’t, it just takes a little more work on my part to figure out how to ask Recorded Future more of what I want to know in a way it can understand and answer back to me in the expected format).

I’ll write a future post on Recorded Future when I can get it answer a question the way I expected it to.

Upon reflection, The Recorded Future is a Query Engine that you ask questions to, it will then go and retrieve the data over time (about future drug announcements, future legal issues, future appearances, as a series of alerts.  Once the data is collected, it can be charted as I’ve shown above, so to be fair to Recorded Future, I should go back in a few days and look at what it collected about Tiger Woods or Angelina Jolie.