Summary of meeting on 30th June 2005
Progress update on the php inference engine and parser (both complete) and minor bug in prolog prototype (justifications appearing twice in the IN-LIST)
Getting a timetable online is a high priority since re-implementing leads to a lot of work
Write the document demonstrating possible inferences in parralell with implementing the jtms
Discussion on possible related areas where inference could be useful for example in the case of Significance and Validity a flaw in the methodology may raise questions on the significance of the research. There is a similar relationship between originality and significance.
Relevance seems quite orthogonal but could have particular conferences or papers where part of the paper is relevant but it is so significant it should still be published.
Study the hypothesis carefully since this is in many ways the heart of the paper, is there a clear claim?
presentation and hypothesis loosely related for example is the hypothesis clearly stated
Potential Inference Backbone is hypothesis? => is evidence? => is methodology sound? => does analysis support hypothesis? => is significant?
Presentation can always be fixed so should be considered quite minor in comparison to problems with originality, validity or signifiacance
To demonstrate jtms it would be nice to have something with multiple justifications so when one is withdrawn the inferred outcome can still stand
Relevance Example For a conference on machine learning we can accept symbolic or sub-symbolic entries ie/ genetic algorithms cound and so does decision trees
Papers with multiple hypothesis that all support significance could be used so if one hypothesis has a flawed analysis then the paper could still stand on the basis of the other hypothesis.
Similarly multiple domains could be used so weakening the significance in one domain could still allow the paper to stand on the basis of the others.
Next meeting 7th July 11am