Expertise profiling comes up quite a bit with our customers. There is a lot of opportunity to streamline the painful email interactions (and over-the-cubicle-wall interactions) where people asking the same question many and/or don't know who to ask. Typically, our customers want to:

  1. Non-obtrusively route questions to the right people

  2. Notify relevant users of events in the system

  3. Quickly connect users with similar interests / skill sets

  4. Provide a level of confidence that an expert's responses are legitimate

To date, our ranking of expertise in Forumshas been based on what people say is their expertise and their points from answering questions (or providing assistance). Moving forward, there are a lot of opportunities to broaden the scope of expertise measurement.

 

To start with, calculating expertise can be done through implicit and explicit profiling. Explicit being the profile data that is actively managed by the user and/or company (e.g. "I am an expert in C# and a novice in Swing"). And implicit being the system's aggregate understanding of expertise based on the content and interactions provided by that user (e.g. Bill has 590 points in "Business Law" and 789 instances of "employment agreement").

 

How to Calculate

 

There are a variety of ways to calculate both explicit and implicit expertise, but the goal is to bring together the most relevant measures and provide an overall score that can serve as the basis for the intelligence described above (where to route questions, who is notified and when, etc.); however, the weighting of the different forms always depends on the use case.

 

The main areas are:

  1. Company Assessment: What does the sponsoring company say about the user's skill set (Explicit)?

  2. User Assessment: What does the user say about their own skill set (Explicit)?

  3. Certifications: What external accreditation does the user have (Explicit)?

  4. Interactions/Content Analysis: What does the system understand about the user's skill set based on interactions and content (Implicit)?

  5. Community Endorsement: What does the community say about the user's expertise (Explicit)?

  6. Points: How has the user scored in different areas of the community (Implicit)?

Summary

 

The most common use case is the routing of questions to people with the appropriate skill set based on the aggregate profile (a much more efficient mechanism for resolution than email). However, an understanding of expertise can be used in many ways, such as making social networking easier, using as the basis for incentives, improving personalization, etc. Right now, our system only deals with 2 and 6, but we are working on finding more advanced ways to calculate expertise as well as what to do with it when you know.