KYOTO Logo
Knowledge Yielding Ontologies
for Transition-based Organization
  • Increase font size
  • Default font size
  • Decrease font size
KYOTO Project Forum
Welcome, Guest
Please Login or Register.    Lost Password?
Re:D6.4 Automatic Deduction and Inferencing (1 viewing) (1) Guests
Go to bottom Post Reply Favoured: 0
TOPIC: Re:D6.4 Automatic Deduction and Inferencing
#38
AmandaHicks (User)
Fresh Boarder
Posts: 9
graph
User Offline Click here to see the profile of this user
D6.4 Automatic Deduction and Inferencing 2 Years ago Karma: 3&nbsp&nbsp
D6.4 Automatic deduction and inferencing techniques (software) is due on August 31.

What are we supposed to do for this?

We need to have a detailed discussion with the whole group regarding how much inferencing to do and what kind of inferencing needs to be done.

There are three parts to this issue.

1. What are the kybots doing? What do they need? What kinds of conditions, constraints?

2. What kind of facts are we producing? This may require producing another system (or Kybots _layer_-2) that combines information and verifies it (prior to reasoning with it? For instance, do not want to reason with inconsistent facts.)

3. The ontology itself
a. Can we somehow define how inferencing is done on the ontology in a generic way? E.g., Transitivity is a kind of mechanism, but do the properties inherited along subclass differ from the properties inherited along part_of?
b. In terms of the gear wheel model, where will various bits of information, i.e., axioms, instances of relations, etc, be stored and or swapped? Where are the rules of inference and general axioms stored?

Proposal: we define a simple inference engine that takes as input:
i. configuration file that defines per relation:
1. relation _meta_ properties such as transitivity
2. what properties are inherited through transitivity
ii. an ontology

For example, subclass and part are transitive Sumo relations, subclass inherits all properties, part only inherits functions or roles. If you then use SUMO for Kybots, the engine will apply inferencing to the KAF patterns to validate semantic constraints. If you want to use DOLCE or the KYOTO ontology, you need to define the _meta_ properties of relations and the engine should be able to do the same with this ontology to the same KAF.

We need to investigate the available inferencing engines and see if we need to build our own (simpler but more efficient and generic). Posting references for such an investigation is a welcome contribution!
 
Report to moderator   Logged Logged  
  The administrator has disabled public write access.
#40
PiekVossen (Admin)
Admin
Posts: 9
graph
User Offline Click here to see the profile of this user
Gender: Male Birthdate: 1960-10-24
Re:D6.4 Automatic Deduction and Inferencing 2 Years ago Karma: 0&nbsp&nbsp
I would like to start this discussion by looking at some specific examples. We are looking in the English sample data_base_ and try to work out a few cases.

This involves:

1. manually look for the KAF structures that relate to the same fact
2. simulate the output of the Kybot as now proposed by German
3. define a representation for the final fact
4. define what kind of steps are required to go from 2. to 3.
 
Report to moderator   Logged Logged  
  The administrator has disabled public write access.
#45
AmandaHicks (User)
Fresh Boarder
Posts: 9
graph
User Offline Click here to see the profile of this user
Re:D6.4 Automatic Deduction and Inferencing 2 Years ago Karma: 3&nbsp&nbsp
These comments are from Christiane. I am posting them here since she is having trouble posting to this forum herself. Just a reminder that alll members of WP06 are especially encouraged to participate in this thread since it is about what we need to do to produce our next deliverable. Cheers, Amanda

Thanks for asking the inference questions. I think we had talked
about this before--it's all a bit mysterious. What exactly
is meant by inferencing, if anything above and beyond reasoning over the
WN/ontology relations? (I.e., if a blackhorn sheep is
a kinda sheep and a sheep is a kinda mammal, the we can "infer" that a
blackhorn sheep is a kinda animal...same with
parts etc. Trivial for people but very useful, of course, for automatic
document manipulation. What else is planned and needs
to be done by Berlin?)
I hope we can discuss this later in May.
 
Report to moderator   Logged Logged  
  The administrator has disabled public write access.
Go to top Post Reply
Powered by FireBoardget the latest posts directly to your desktop

ICT-211423 - 2008 © Kyoto Consortium