Forum OpenACS Q&A: Knowledge Management System

Posted by John Cusack on
How can i use OpenACS framework to develop an Ontology-based Knowledge Management System?
Posted by Jeff Davis on
Via installing OpenACS and then customizing it :)

Seriously, that's a pretty broad question and I think the answer could fill multiple books. Maybe you should give some more specific examples of what you are trying to do and we might be able to tell you what parts of OpenACS would be useful and what you might need to change.

I am working on some KM related stuff currently which might be useful but it is a ways from being delivered.

Posted by John Cusack on
:),Thanks for the previous answer, for your information, i'm a student working on a KM project with just a brief idea.Basically, we would want to have a knowledge repository to store documents, procedures etc. for future reference. Then, there should be a retrieval engine right?The project team has decided to use ontology to represent the knowledge structure. So, i would to know more about developing a retrieval engine that use ontology as a knowledge structure, on what it will needs, how will it be and related concepts in order to implement it using openACS. :) .Looking forward for any assistant.
Posted by Frank Bergmann on
Hi Mohd,

we have got a KM research group working at the La Salle university in Barcelona, Spain with a total of 4 final thesis students. We are working currently for two customers, a small strategic consulting company and the HP Professional Printers division.

For the consulting company we basicly develop a new filestorage module with template documents for each project phase.

For HP we develop an "Expert Finder" module, which is basicly a search engine, where we check for the authors of the documents that match a search query. This project involves the classification of keywords, persons, documents and workgroups into an ontology, in order to restrict the search space and to desambiguate "names".

Name disambiguation is especially important, because names can refer to many persons, but it is vast majority of the information we've got. So if have got users classified into the ontology, and we've got a document classified, then we can determine the probability that a name in the document refers to one user or another.

We use the probabilistic calculus as base for our system, with ontology classification represented as a vector of probabilities, one probability for each node. This representation is very database-friendly, even though the size of the ontology gets somewhat limited.

Does this match a bit with your ideas?

We've got a lot of experience with description logics and frame-based symbolic systems, but we are going now for the probabilistic calculus, because it is closer to the scoring algorithms used in Internet search engines while providing sound semantics.

It would be definitely be interesting to see if we could find common ideas and or common projects.