• Teleconference to discuss Causal relationship investigation (1)

Teleconference to discuss causal relationship investigation

Date: 2010-02-16

Participants

• Siemens: Markus, Florian, Volker

• ?OntoText: Vassil

• ?AstraZeneca: Peter, Gunnar, Mats, Bosse

Notes

The call objective was to discuss how Causal relationship Investigation can be applied to help scientists answer complex questions (in dialogue with the "application").

The initial query tested by ?OntoText raised excitement and all participants agreed this could be extremely valuable for the use case.

Peter proposed a natural language representation of the results. This would give the scientists a good chance to easy prioritize and understanding the result hypothesis.

Volker wanted to investigate more in depth how Machine Learning can be applied. Two types of machine learning; concept-level analyses (gene X causes disease Y) or instance-level analyses (in X% of the test we found conclusion Y) is in scope.

Gunnar don't see the need to distinguish necessary, sufficient, contributory cause, however time relation may be important.

We agreed to develop a plan (peter to take the lead) with multiple steps/approaches (iterative and start simple) 1) brute force check with expert; 2) simple heuristic and cutting; 2) uncertainty reasoning based on statistics While the plan is developed ?OntoText and AZ will continue to evaluate queries to identify most important concepts for this work. ?OntoText's role will be to provide necessary tools and infrastructure. Siemens need to analyze the problem.

Resources (Siemens) spent on this effort should be reported to wp3 if is it applicable to machine learning.

We will need face2face or video conferences to better understand LLD, Scientists requirements and ML opportunities. Will address this need in the plan.

To be able to present some result in next review we have to deliver first step/iteration in June.

Actions

• ?OntoText and AZ scientists to evaluate a couple of more questions.

• Siemens to write a two page input that analyses the problem describing how Machine Learning can be applied.(Due date 2010-03-09)

• ?AstraZeneca (Peter) to write plan for the work. (Due date 2010-03-09)

Next meeting

To be decided