SAB Meeting Notes, Feb 26th 2009

iAct

* investigates the adaptive visualisation of large displays
* uses stereoscopic projector for 3D display
* combine 3D with high-resolution 2D text
* evaluation: game playing, hot wire game, Poggendorf illusion
* results: dimensionality does not affect retrieval, but display size does, reproduced evidence of advantage of procedural memory (Handlungsgedaechtnis)
* "collaboration" is modeled as a group view generated from several private views
* the project organized a workshop. They cooperate with Lufthansa, Fresenius Kabi, Microsoft.

Proposal for Capture (Stricker)

* 3D scene reconstruction from images
* tracking of user's hands etc. using only one spherical camera
* use portable head-up display to augment reality with virtual objects
* do discrepancy checks (CAD data vs. captured images)
* approach: use high quality pictures (lots of data) for image reconstruction (amount of data challenging)

1st Workshop on Context Management

* a day workshop that linked projects Carmina, Perspecting, Take
* also provided repository, website, mailinglist (trac.dfki.de )

MyMory

* investigate context-related work, represent/elicit the user's context (re-contextualisation)
* exploit context for information retrieval
* attention as indicator for relevant context aspects
* MyMory investigatese text documents, technically via annotations and a diary study
* System with 4 level architecture, (i) platform level (e.g. ontology), (ii) service level (e.g. context services, information retrieval), (iii) application level (e.g. Wiki), (iv) appliance level
* Platform was used by other projects at DFKI, too (collaboration developed over the timespan of three full projects)
* attention evidence is collected via a webcam, eyetracker and markings in the document
* implements activation spreading ("entailment spreading" ~ inference)
* represent short term/medium term/long term context
* attention-oriented retrieval engine
* Wiki: manual and automated annotations
* coupling with DynaQ (other project in the department)
* dashboard application: visualise timeline
* experiment: give subjects a task that they have to perform using only a MyMory recording
* experiment: classification of contexts (only used themselves as subjects)
* reported improvements to their relevance feedback and ranking
* performed industrial case studies with Asstel (insurance) and Siemens
* one project member spent a reasearch internship at Microsoft ("scientific exchange")
* i-ducator: detecting of reading modes via eyetracking, published in SIGINT
* Questions: evaluation

Prospecting

* report on the first six weeks of the project
* employ user attention for situated retrieval
* ontology-based information extraction
* represent search experiences/search history
* architecture makes use of attention data, employs EEG and eyetracker
* uses memory model/mental model (long term attention history,
personal information model, information item store, seeking process store)
* evaluation: measure precision & recall
* novelty/understandability extimation
* until now: project team and infrastructure have been established
* Questions from panel: (i) what about privacy of user's data? (ii) EEG - they hope to see relationships between brain activity and relevance of documents, (iii) they collaborate with 'Bremen'

Feedback to my talks

* Check cognitive load theory
* Check whether iterative logic programming (or something like that) allows to usefully combine rule sets (proposal by Paul)
* be more specific about student's mastery, more detailed modelling of mastery (Dimitra)
* slide about assertion level proving - do not use "C" to label the conclusion
* Research question: does knowledge about granularity help student learning? (Erica)
* Need to justify my features (via theories?)

Anonymizing References in LaTeX

The problem: submit a TeX paper where self-references are removed (of course, removing them completely from the source will make it more difficult to create the camera-ready later).

Step 1: duplicate your entries, and attach a suffix (e.g. "-orig") always to one of the copies. Use that one to create the camera-ready later.
Step 2: Take the other copies, and remove all the information except the year of publication. As the author, indicate "\zzz{} ...", where \zzz is a macro defined in the main tex document (\newcommand[0]{\zzz}{}). The \zzz will help to make the left-out references appear at the bottom of the list.

Now each entry is dublicated as:

\inproceedings{paper01-orig,
author = {Real Author},
booktitle = {Real Book Title},
year = {2010}
}

\inproceedings{paper01,
author = {\zzz{} ...},
year = {2010},
note= {$<$ hidden reference $>$}
}

In the submission copy, all references indicate ... as the author. To enable references again for the camera-ready, rename the labels again (e.g. replace "-orig" with "").