LarKC WP4+WP6 Teleconference on Reasoning with Noisy Data, 11.00 -12.00, Sept 14, 2010
The skype ID of the chair: zhisheng_huang
Participants
Zhisheng (VUA), Gaston (VUA), Emanuele (CEFRIEL), Stanley (Saltlux), Daniele Dell'Aglio (CEFRIEL)
Sources
all triples of Road sign management usecase at: http://wiki.larkc.eu/LarkcProject/WP6/WorkInProgress/OSM
Stanley's PPT: http://svn.larkc.eu/wp4/wiki-resource/presentations/2010-09-14_Reasoning_with_noise_geo_data.ppt
Discussions
Zhisheng: There are two strategies for reasoning with noisy data: one is forget-only strategy in which we can make the reasoning by ignoring the error data, another is correction-involved strategy in which we create some new data for making the correction reasoning. For the latter, we have two different approaches: one is the newly created data are added into the data set, which can be re-used next time, another approach is the new data are created for this special query, without adding them into the data set for re-use of other scenarios.
Zhisheng: There is another dimension of noisy data: uncertain data, incomplete data, and inconsistent data. We are going to classify all of the cases of noisy data in Stanley's PPT, based on the dimensions above.
- Emanuele Della Valle:
- case 1: incomplete data (missing juction)
- case 1 : uncertain data (is there a bridge or not?)
- case 2: inconsistent arising from data integration (a road cannot be inside a building)
- case 2: uncertain data (is the road or the building that is wrongly placed?)
- case 3: data duplication (road 2 has be drawn on top of road 1)
- case 4: missing data
- case 5: incomplete data
- case 5: easier to connect because the two roads have the same name
Zhisheng: The question about case 3: is it forbidden or allowed, based on the Korean road regularities?
Stanley: Yes, it is allowed, however, it is bad for reasoning
Gaston Tagni: http://www.openstreetmap.org/
Emanuele Della Valle: For case 4, the idea: use google or Korean Road Sig to find evidence that the link actually exists
Gaston Tagni: it seems that the Google Maps API does allow for the kind of query Emanuelle mentioned: http://code.google.com/apis/maps/documentation/directions/
- Actions:
Stanley is going to provide the test data sets + some test queries : http://wiki.larkc.eu/LarkcProject/WP6/WorkInProgress/OSM
- Gaston and Zhisheng are going to discuss how to make the tests to get correct answers.
- Zhisheng: We are going to develop a framework of reasoning with noisy data, not only for Korean road sign management, but also for other cases.
Next teleconference on Reasoning with Noisy data: In three weeks. Namely, Oct 5th, starting 11:00.