Tacit Knowledge Mining

1
Tacit Knowledge Mining

• John CannyComputer Science DivisionUC Berkeley

2
Motivation

• People often collaborate across geographic,
  cultural and disciplinary boundaries.
• Workgroups are more short-lived today. There is
  less time to learn each others skills sets, to
  develop trust, and a voice within the group.
• On the other hand, there is a wealth of
  information available in groupware systems about
  peoples interactions with each other and with
  data.

3
An opportunity

• We know that much of the useful knowledge in a
  group or organization is tacit. Can we recover
  and use tacit knowledge?
• Human knowledge and knowhow is difficult and
  expensive to codify.
• But the knowledge encoded in activity data may be
  less so. Examples expertise discovery, group
  structure, document dependencies.

4
Platform Lotus Notes for now

• Data sources
• Direct messaging sender, receiver, time,
  duration..
• Topical discussions or threaded email.
• Document access (by links or directories)
• Document search (by text queries)
• Annotations on documents.

5
Analysis methods

• Social Networks. Centrality measures for
  estimating authority and prestige.

6
Analysis methods

• Clustering. Discovering tacit groups, and related
  sets of documents.
• Classification. Use a knowledge hierarchy to
  classify documents, and compute expertise
  profiles.

7
What we want to assist

• Expertise discovery. Expertise profiles include
  authority in each area.
• Document access. We compute document authorities
  and use them in ranking.
• Group dynamics. Does the communication network
  contain any problematic structures?

8
What we want to assist

• Sensemaking, document context. Record the
  document creation process. Prioritize records.
• Attention management. Infer a current task and
  adjust awareness of documents and other people
  according to their proximity to the task.
• Perspectives. Organize annotations (when
  available) on a document according to the
  expertise of author and annotator.