Expertise location: linking social networks and text mining


A very interesting article in the Guardian today, US military targets social nets, describes new expertise location technologies.

Expertise location has always been a central ‘killer app’ first sought by knowledge management and now part of the promised of Web 2.0. It is a fundamental driver in any large organization being able to tap its own capabilities and take advantage of being large. This was always epitomized by the quote from Lew Platt, who as CEO of HP famously said “If HP knew what HP knows, it would be three times more profitable!”.

I wrote in 2005 about how Morgan Stanley was finding that blogging was trumping in effectiveness its years of efforts into dedicated expertise location systems. The next layer is tapping social network and content creation patterns to identify experts, as has been implemented in some content management systems (CMS) over the last couple of years. This can be taken further when used within online communities and social networks, as SRI International is currently doing:

SRI International based in Menlo Park, California, teamed up with military officers to build a new social analytics tool called iLink that generates models and helps streamline the process by which a specific expert in an online community can be found.

In simple terms, iLink is a machine learning-based system that models users and content in a social network and then points the user to relevant content.

The iLink system had several goals, including real-time learning by matching queries and communities users; adapting to user demands and directions, providing accuracy in message targeting and routing and, finally, dynamic user profile correction based on community behaviours and identification of community experts.

The learning in iLink occurs by watching a natural social network, and selecting effective strategies that emerge from the system as the members try to solve problems. The system continuously monitors the real social network and it is capable of drafting from the social network’s learning.

The iLink software uses artificial intelligence software and message routing technology to help the system learn about the online participants and move specific questions to those who are best equipped to answer them. The SRI scientists basically build a profile of each person in the community and the iLink system starts to learn about the movement of information around the community.

In my post last year on the state of social networking software for the enterprise I did a review of some of the platforms at the time, including the application to expertise location. Once business activity starts to shift to social networks, and the way people create and consume content can be better understood in a social context, many new applications emerge. Microsoft with SharePoint 2007 is trying to get into the expertise location market, while IBM’s latest versions of Lotus Connections are shifting into more sophisticated approaches to finding expertise. But much of this will be driven by cultural as well as technological efforts to break down silos.

It is turning out that one of the greatest benefits of a shift to the use of social networks in the enterprise is making talent more visible and accessible. Intelligent text mining is taking this to the next level. There will be some massive value creation in this space.

2 replies
  1. Samuel
    Samuel says:

    Great post. I agree that using ‘intelligent text mining’ is an interesting approach to expertise location in companies (and on the internet). We experimented with this some time ago in our company with interesting results. This experiment was set up because – as we all experience – employees fill in their Yellow Page profile, but don’t keep them up to date. Relating the filled-in profile to mining could trigger employees to keep it up to date. And it could also (partially) fill in their profile.
    We also combined this with a more social approach, which is now being capitalized in Guruscan(.nl). Because using mining to find and define expertise limits you to what’s in databases. And when we write reports about a tool, we don’t mention we’re very good at PERL programming for instance. So, this social layer collects the tacit stuff.
    Here are some references:
    – Samuel Driessen, Willem-Olaf Huijsen, Marjan Grootveld, “A framework for evaluating knowledge-mapping tools”, Journal of Knowledge Management, 2007, Vol. 11, Iss. 2, page 109-117.
    – Willem-Olaf Huijsen, Samuël J. Driessen, Dion Slijp, “ExpertFinder: Collaborative Expertise Localization”, I-Media 2007.

Comments are closed.