Hildeberto Mendonca
|
Posted: October 14, 2008 20:34 by Hildeberto Mendonca
|
|
One of the forms to infer meanings extracted from modalities data is defining and processing rules. A rule is a good instrument to represent assumptions, mainly when we have a domain of application. This topic may stimulate the discussion about the support of rule engines in the meanings4fusion project. You are welcome to contribute! |
Rule Engine Support
4 posts
Replies: 3 - Last Post: November 23, 2008 11:55
by: Hildeberto Mendonca
by: Hildeberto Mendonca
showing 1 - 4 of 4
OlgaVybornova
|
Posted: October 19, 2008 20:26 by OlgaVybornova
|
|
There are numerous methods that can be applied to derive meanings, semantics of events, to discover similar events, to predict events, to understand the context. The crucial part of the decision-making process is contextual reasoning. We are looking for the best, the most efficient solutions for this problem - going from ontology and knowledge representation tasks we want to make the rule engine unit deal with parallel processing of multimodal data streams, set priorities of fusion rules depending on the context, apply temporal logic, spatial reasoning, partial and dynamic ontology mappings to improve communication between agents. Well, if anybody could share his/her valuable experience with different applications and describe what methods you used, and what were advantages and drawbacks, it would be great! A compressed discussion of practical experiences with semantic data integration can be more profitable for all participants than reading tons of papers on each topic separately. |
|
Hildeberto Mendonca
|
Posted: October 20, 2008 15:59 by Hildeberto Mendonca
|
It's important to emphasize that we are looking for the best solution, of course, but if it is open source will be even better  . We are motivated to contribute with a third-party open source project if it is suitable for our needs.It's true that is very difficult to find a silver bullet out there, but when we are working in a quite new technology, integration stuff is also a very exciting work .
|
|
Hildeberto Mendonca
|
Posted: November 23, 2008 11:55 by Hildeberto Mendonca
|
|
I think we are close to a final decision about the Rule Engine Support. We don't have a final decision yet because our investigation was a little bit frustrating. We haven't found many options and the available ones have some restrictions that is not compatible with this project. According to all comments I've read, the best rule engine available is ILOG JRules, which is a very expensive product, but very complete. Its main competitor is Fair Isaac Blaze Advisor, also expensive and complete. Not so expensive, but with an amateur attempt to be serious, there is Jess, a small, light and fast rule engine. The problem with all rule engines above is that, none of them can be distributed within a open source project. None of them are open source or have a special partnership or appropriate license for open source projects. Jess can be licensed free of charge for academic use, but we need people outside of the academy too, like most of you. So, there were summarily discarded. On the open source world, we found only two good options: 1) SOAR, a general cognitive architecture for developing systems that exhibit intelligent behavior. Its license is BSD, which gives complete freedom of changing and reusing. The problem with SOAR is its complexity. It tries to be generic, solving many kinds of problems through different approaches, and it is also big, difficult to embed in an application. 2) Drools, an enhanced rule engine implementation. It is totally written in Java, which makes it totally compatible with the Meanings4Fusion implementation. We can embed and distribute it in our releases because it has a flexible open source license fully compatible with ours. Drools is a JBoss sub-project, assuring a large community around it and its long life as an open source project. With Drools, we will have a permanent support of a rule engine. So, the priority now is to integrate it as soon as possible. However, people might be good users of other tools, like the ones mentioned above. So, after the Drools integration, we are thinking about an adapter to connect Meanings4Fusion with other tools, but it is a plan for 2009. The next release (0.6), planned for December 5, you will have a complete idea about the role of rules in our system. We hope to count on you during this rich experience. |
Replies: 3 - Last Post: November 23, 2008 11:55
by: Hildeberto Mendonca
by: Hildeberto Mendonca
