Mopsos

"If only we knew what we know…" When top managers become concerned by their lack of visibility of the company’s collective knowledge, they start a «corporate yellow pages» project to improve the online corporate directory, which usually only gives business card informations. In a typical top-down approach, taxonomies of expertise are created, on line résumé forms are designed and sent to all employees to fill, and some search engine is put on top.

Back in the mid to late 90s, some companies like BP and Schlumberger have been successful with this top-down approach, others much less. Actually, it proved to be more expensive than anticipated. Those who ran the show had to provide incentives and rewards to solve two key quality problems:
- making sure that the taxonomy of expertise was actually simple enough to be filled by users and yet useful enough to locate experts
- making sure that the data were regularly updated

Many other approaches have been experimented afterwards, but I haven't read anything that looks like a white paper of good practices regarding Corporate Yellow Pages.

Here's how mine would go. Anyone to write it with me?

Starting from 2000, semantic analysis of free text was introduced as an alternative. Tacit Software pioneered this approach using corporate e-mails as the basic material to mine, and implemented its solution to companies like Aventis and Northrop-Grumman. (Note: The Fraunhofer Institut was also working on something similar at the time. I wonder what happened to the project…). Other semantic engine technologies like Autonomy or Entopia started to be used in a similar way using whatever publication made by the person as the basic mining material. This turned out to be helpful but insufficient. Words defining expertise are context-specific. Thus expertise cannot be easily captured by a list of words. Technologies like Arisem’s went one step further, allowing the use of different ontologies to map different fields of expertise, thus allowing different viewpoints to mine an existing body of text. But in any case, such tools require a large quantity of authored free text to start with, which raises issues about peoples' writing habits (some never write anything, especially managers ;-)), and legal issues regarding privacy as well, especially in Europe.

Business-oriented social networking tools like LinkedIn, Ryze, Spoke, ZeroDegrees came next in 2002-2003. They can be considered as offsprings of dating service sites like Match or Meetic in France. They focus on the links between individuals rather than on the individuals themselves (a.k.a. the ties of the graph instead of the nodes). They appeal to the users because they are user-centric and open-ended. They are not about matching one particular individual with a question with another one with an answer, but more about connecting people based on affinities and domains of interest. They are clearly attracting most of the attention today, and raising capital as well.

If I were to take the lead of a yellow pages project in a big company, I would use a combination of the three approaches depending on the underlying business problem:

General business networking –
Ask people to register on a professional social networking tool like LinkedIn (US). I would also suggest for security reasons other places -and other countries- to register with like OpenBC (Germany)... This is free. It is likely that professional societies and associations will implement such tools within their network if they aren’t doing it already. Register there too.
Similar patterns of business networking can be useful inside large international companies, especially to obtain new connections to clients and prospects. An enterprise edition of these tools might be useful here.
Note : It is likely that the basic professional informations regarding individuals like name address, resumé etc. will reside on a private secure server in the near future so that all social networking and expert location tools will index the same and unique file. This is obviously necessary to ensure quality and freshness of data. We need some sort of standard XML format to do that. Maybe it’s already there (vcCard, FOAF ?). Can anybody help me here?

Free text based expert location -
This works best for writing populations that do not limit themselves to e-mail. This would involve asking people to upload on a private space on the company server a selection of documents they wrote that are most representative of their expertise (great for bloggers !). Then index the corresponding databases and run the clustering application of some semantic engine.
Note : if the company is using ontology-based tools like Arisem, each community of practice of the company should be in charge of defining the ontology of their domain, which would then be used by the entire company. This was the approach successfully implemented by Jean-Jacques Régnier for CNES (French National Center for Space Studies).

Top-down expertise directory -
Establish a simple taxonomy of expertise for the company. It has to be simple and easy to fill - 15 minutes max. The list of active communites of practice of the company is part of this taxonomy, because the CoPs people belong to are good indicators of their area of expertise. At a deeper level, each Community of Practice of the company should be empowered and asked to elaborate the taxonomy of expertise -if not its ontology- of its knowledge domain, which could be then consolidated at the company level. The more communities of practice an employee belongs to, the more his/her expertise could be codified and easy to locate.
The corporate directory is finally equiped with a search engine sitting on top allowing users to search on whatever combination of fields. A CBR technology like Kaidara comes in handy here (It works great!).

Integrating the three in a single application a la Gurunet would be the "final" step. It could include goodies like automated alerts on my e-mail client whenever some employee with a given declared expertise appears in the database.

No-one has commented on this entry (yet).

Thanks for signing in, . Now you can comment. (sign out)