I am attending the ESSIR Summer School in Padua this summer. I will be trying a bit of live blogging during the summer school. What I will be noting down, is my view on the presented topics - Using the presentations in my context/research.

First live blogging is about Introduction to Information Retrieval presented by C. J. ‘Keith’ van Rijsbergen. The topics in this presentation are What is IR? IR Models, Precision and Recal, Relevance feedback. Here I will focus on what Information retrieval is.

Keith presented this model of the fields scientific paradigms

Scientific paradigm

Theory

<->

Experiments

<->

Real Life

Top down, and Bottom up -Approach

My research is clearly the Bottom up approach. I start with real users, and real (as well as simulated) task, experiences and from there on I will try to develop theory.

What is IR?
Generally speaking IR is about retrieval of textual documents. As a former student of Peter Ingwersen I would like to add that interactive - the interaction between the user and the system also must be present. I would say that for me, there is no point of studying IR without studying the users.  But I can see that studying/investigating algorithms and IR-systems can be made without users. But at some point user studies must be made.

‘Retrieval systems count things’

Systems count words locally in a document/object, and then they count words globally/external from the document and violá.

- The strength of the field of IR is evaluation

How do we evaluate the invented IR-systems. Demonstrate that the systems work. Not just demonstrate the speed. Measuring what the IR system returns in connection with the user. This will come up!

Here are some highlights from Keith’s presentation

- 3 dimensional experimentation is when you add the users context - do that

- Controlled variables in experiments are test collections, topics, assesments (from Kal Järvelin)

- Use of words are problematic - words are to ambigius

- In real life - you don’t have relevance in advance

And my favorite from Swanson (1988):

- A document relevance depends on other seen documents - relevance is not indeendent

References
Swanson (1988). Historical note: Information retrieval and the future of an illusion. In: Journal of the American Society for Information Science. 39 (2). p. 92-98.