E-cognos project’s final meeting in Paris¶

E-cognos is a European Union-funded project, consisting of 4 end-users (big contractors and the like) and 2 research companies. They've done research on knowledge management in the construction industry. On 30 september 2003 they hosted their final meeting in Paris. I've written down part of what I heard and mixed it with a bit of my own take on it. If you see something you don't like, you can choose who to blame, me or them :-) Most of my remarks are indicated by "I think" etc, so it 'll probably be clear.

In the introductory speach they made the nice remark that knowledge management = making the implicit explicit. When dealing with knowledge you have to deal with various levels, going through and through the various levels over and over again:

• Creating knowledge.
• Identifying knowledge.
• Collecting knowledge.
• Organising knowledge.
• Sharing knowledge.
• Adapting knowledge.
• Using knowledge.

Keep this in mind. Omitting a few will give you a headache. If you don't share knowledge, you will get less knowledge from others in return. If you don't adapt knowledge, it 'll be less useful. Etcetera.

Good to hear was that the software will be made available for free (GPL license) somewhere in october. CSTB made a lot of it. They didn't do a presentation so I mention them here. You see this more and more in European projects, which is a good thing imho. Most of the eConstruct stuff was also available (see bcxml.net for some of it).

Social pocesses of information retrieval (university of Salford)

"In the past, you asked friends for an answer, now you ask google". Times are a'changing.

If you need information, you tend to look in four places. The idea I got from the presentation is that you walk through the four of them in sequence till you find an answer.

1. Your own knowledge base (personal experience, stored files). Note: for me, google is almost an external memory. So I guess that for me, it fits in this category. For many, it will be in number four.
2. Knowledgeable colleagues (internal).
3. Knowledgeable colleagues (external).
4. Proprietary/public sources (internet, subscription CDs, books).

There is a focus in e-cognos on "communities of practice", which is a thing that's happening all the time. E-cognos uses standard things like user profiling to try to form those communities automatically. E-cognos then also tries to pro-actively search out good information for specific kinds of users.

Collaborative approach

When you search for information, there is a definitive scale from a structured search to an unstructured search.

• Consulting a document management system. Very structured. You look for a specific document.
• Also structured is using a search engine (like google). But among the results you see other things that might interest you and you wander away a bit from your structuredness...
• In (email) discussion groups, you tend to brainstorm. Coining ideas, discussing a bit, listening to other's ideas. This is mostly unstructured, but at least you will probably find many new, interesting ideas.
• Even more unstructured is bibbling (neologism...). Browsing through libraries and bibliographies. This way you can even find a new knowledgeable community.
• Most unstructured is the serendipity level. (Serendipity is making fortunate discoveries by accident).

E-cognos tries to supply all these levels in their "portal".

Global architecture overview

Boiled down to the core, e-cognos provides a few services (like an ontology server). On top of this, they added an interface/API by which to access it. You can access it using web services - by which they mean SOAP.

I personally moderately abhor SOAP. Most of what SOAP does can be done with simple HTTP and XML. Saver, cleaner, better. SOAP, to me, is hyped by vendors trying to sell their big toolkits. You can do the same with the basic http/xml support that is available in every programming language...

The idea I got from the presentations is that the biggest thing in this e-cognos architecture is the ontology(-server). They created/converted a few ontologies (in this case: hierarchical sets of keywords with translations and synonyms). Added to that is a bit of user profiling etc. I don't have a full picture of it, but the documents will be on their website somewhere late october. Probably the ontology server is the most visible part, there's definitively more "under the hood".

Presentation by OTH / Derbi (france)

They integrated the e-cognos facilities into their existing document management system. This means that you can use the ontology's list of equivalent terms, synonyms, broader/narrower terms to refine your query.

When a new project in Derbi is started, you can now look for similar projects and relevant experience.

Their experience was that the e-cognos API is good enough to integrate with an existing document management system.

YIT (Finland)

(Biggest contractor in Finland).

They're using product models and EDM (express data manager) big time. Mostly they use archicad to fill the product model.

They showed a slick demo about exporting an archicad model to their own edm/ifc product model ("cove"). They coupled the e-cognos web interface with above "cove" application. This way you could use e-cognos's keyword functionality and the user profiling functionality to search for data in the model and to send relevant information to the people concerned about that information. It looked pretty well integrated.

They showed a scenario where slabs weren't the correct thickness. Using e-cognos, they found the documentation containing the correct measurements and they found the people that needed to deal with this problem.

In my opinion, the e-cognos interface still looks a bit rough, it needs a usability expert to go over it once. But it's not bad at all. You can see they spend a great deal of time building all that functionality.

What YIT learned out of this was:

• Social processes are the key. Finding the correct people, linking people, providing people with the relevant information for their point of view, not showing piles of irrelevant documents, etc.
• The e-cognos tools really provided useful knowledge in the day-to-day work in the company (for those people that tested it).

Hochtief

Big German global contractor.

Quote: "incomplete info has little value".

I smiled broadly when I saw that they had used zope and plone. They used it as the web interface of their existing content management systems. It also was able to connect to the e-cognos infrastructure. I'm a big fan of zope and plone (and python). It was nice to see the Hochtief guy being enthousiastic about it.

Their use case was the process of "structural design calculation". Dealing with legal standards (which are in the existing content management system) etc. Lots of information and the need to find experts. Just the things e-cognos was meant to solve.

Key requirements:

• The need to keep the existing data islands, but with one entry point.
• Good, intuitive user interface. (Fulfilled by using plone/zope mostly out-of-the-box!)

Most of the visible functionality was in the search form. They adapted the standard zope/plone search form to also show the broader/narrower/related terms from the e-cognos ontology so that you could refine your search with them. For those who know plone: they also adapted the "what's related" box to use the narrower and broader terms from the ontology to add a "more loosely related"-box.

The 20 test-users of the system appreciated the following things most:

• Single entry point. Zope/plone as a single entry point for all the existing systems was by itself a big plus.
• Clustering through ontology. Automatically, related items get grouped together. Handy.
• Guided search through ontology.
• Bookmarks. Standard plone feature. If you're a logged-in user you can easily create a personal set of bookmarks of pages in the site. But coupled with the e-cognos ontology this quickly turns your list of bookmarks in your own personal knowledge base. They sounded real enthousiatic about this.

I got the impression that zope/plone + ontology is a powerful combination. Zope/plone seems to shine that way. Great for me, as I'm also working in that direction. Once I've finished a reasonable version of my combination of rdflib and zope I'll email them, easily adding rdf to this might just make it more powerful.

Great job they did.

Taylor-Woodrow

Big British contractor.

Taylor-Woodrow's goal is towards zero defects. To reach this, they need to catch mistakes earlier on in the process. If done good enough, that way there won't be a defect left at handover-time.

Most of the talk was dedicated to a research done for them by a scientist from the Bayswater institute, specialised in integrating the social, human and technical issues in organisations.

E-cognos won't change the business process, but it enriches the toolset of the knowledge workers to enable them to improve the business processes.

Technically, the current ("as-is") position is that the user has to deal with a lot of data sources at the same time. Files on the harddisk. Documents in the content management system. Google. Etc. The "to-be"-position is that the user has a single entry point to all that data.

As-is position organisational/cultural

• Decentralised organisations.
• Consortia-based.
• Local differences ("I use this instead").
• Cultural firewalls ("you won't get my info").
• Competition / blame-culture, which prevents knowledge sharing. You learn most from your mistakes, but sharing knowledge of those mistakes with others might be career-ending...
• There is a lot of knowledge inside people, so finding people is as important as finding documents.

E-cognos has got to earn the right to be used on the floor. Effective use of the e-cognos system depends on developing a knowledge-sharing culture.

From a research by Moffett, McAdam and Parkinson he extracted the distinction between dealing with knowledge management on a technical and an organisational level.

1. There are companies that don't do a thing (40%).
2. Some have a technically sound system in place, but lack the organisational culture to really use it (43%).
3. Others are organisationally ready and willing to really manage their knowledge, but lack a good technical implementation of that will (14%).
4. A mere 3% has both the organisational will/support and the techical system in place.

To-be position organisational/cultural

• Provide access to majority of documents of likely interest. If you can only access 20% of the documents, you still head off to the library immediately before bothering to look in the knowledge system. You need to have most of your documents in place!
• Achieve 80% success rate in finding a specific document (which you know to exist). "I know we've got that ISO 12345 somewhere in our database".
• Google is the "gold standard". Be as user friendly and as useful. It's a tall order...

Conclusion

There is a need for a technical solution, but it has to be completed with the development of a culture of sharing knowledge. And you need trust that you won't be punished for mistakes.

Storing information in the system has to be easy.

Achieving an integrated knowledge sharing system is an evolutionary process. What's needed:

• Senior management support. They are the only once that can take care of the knowledge sharing values (needed to prevent sackings on a lower level, for instance when you admit to an error because you learned a lot from it; you shouldn't be in danger of a being kicked out.)
• Steering committee. Provide sustained and broad support.
• Important: Do a small-scale implementation to establish benefit.
  • Study which specific benefits are sought and also identify other potential benefits.
  • Create a technical system that provides the maximum benefit for the minimal costs. And costs aren't only monetary costs. If adding a useful feature to the system means that everybody spends twice as much time adding documents, then that feature probably isn't worth it.
  • Make sure you can find documents as well as people in the system.
  • Encourage adding existing projects to the system.
• Use converts as advocates.

The end

This ended the presentations. What followed was a brief round of questions.

• There is a difficulty. Company data is often well-protected and there isn't much willingness to share that data with other partners (and future competitors!) in a consortium. How, then, can you share? There wasn't really a fixed answer to this question.
• The knowledge manager has to be the user's friend. That is the required focus.
• There is a real split between ordinary users which get the benefits of this system and the knowledge manager. The second one probably understands what you mean by "ontology", but the user turns pale. Use the term "keywords" for the user.

Nice meeting, handy to get a short overview of the project. It 'll help me to pick out the most useful documents when they arrive at the end of the month. And it will help me when looking at the upcoming open source software they've made.

Regarding the software I'm interested if the additions to zope and plone are part of it...