Re: [Arches] Re: Ideas for GSoC Project: Improve Representation of Spatial-Temporal Relationships
Thanks for all your input! I decided to go with the topic of improving Arches' ability to handle fuzzy temporal endpoints. I've submitted my proposal, and I look forward to hearing your opinions and suggestions. Best, Daphne On Monday, 17 March 2014 18:06:13 UTC-4, Koen Van Daele wrote: Hi Daphne, I think that anything that has to deal with vague temporal or spatial information will automatically turn out te be fairly complex. The fuzzy endpoints of periods is probably the easier one. There the big hurdle is making a good UI that allows for people to enter vague temporal information in a simple and straightforward way that doesn't take to long. We've currently implemented this with a set of widget that allow choosing a start period and an end period whereby both can have different levels of fuzzyness. Eg. A period of time that runs from 'the 8th century BC' till the year 57 BC (Iron Age in Belgium). It's a very textual way of entering data, but that seems to match our users best. Although in actual fact we currently only use the Fuzzy Temporal Intervals for dates of Events (an excavation, a survey, aerial photography, ...) For dating of sites and monuments we use an entirely different approach based solely on a thesaurus of periods. I think the second direction: Allow periods and regions to be associated poses a lot more conceptual issues. I guess one could define a whole set of different Iron Ages, but I have a hard time seeing how it would all work on a practical level. Especially once you start to factor in bringing together data from different implementations and the fact that a region can be vague as well. I also think that another complicating factor can be researches who tend take their chronological system with them. For example, in Flanders, a part of the Iron Age archaeologists use the French division of the Iron Age and another group uses the Dutch division. One group tends to work closer to the French border, the other one closer to the Dutch border. But they also seem to be aligned with the universities the archaeologists graduated from. So I'm not even sure anyone would be capable of geographically defining those ares. Also, when someone then tries to search for Late Iron Age, should they get results from the French Late Iron Age, the Dutch one or both? Or should it be impossible to search for something like 'Late Iron Age' and only possible to search for a specific one? I'm also not sure if the definition of a certain period in a certain area has stayed constant through time. That probably affects things as well. Anyway, it's an absolutely facinating research topic, but one where I think you really need to define a clear scope for your project and resist the tempation of going to far away from it. And make sure that you can communicate to users of your system which choices you've made and why. Cheers, Koen Van: arches...@googlegroups.com javascript: [arches...@googlegroups.comjavascript:] namens Daphne Ippolito [daph...@gmail.com javascript:] Verzonden: zondag 16 maart 2014 20:08 Aan: arches...@googlegroups.com javascript: Onderwerp: [Arches] Re: Ideas for GSoC Project: Improve Representation of Spatial-Temporal Relationships Hi. Thanks for your responses. I don't know Dutch, but your abstract did a good job of explaining what the situation is and the terminology I should be looking into. There seems to be two separate directions this project could go in. 1. Improving the way Arches deals with endpoints of periods by allowing for the input of fuzziness. This task doesn't have a spatial component to it. The project would involve implementing a way for users to specify uncertainty and subjectivity in their period definitions using the algorithms you mention in your post. 2. Creating the ability for periods to be spatially referenced. This task only deals with the spatial, not the temporal, definition of periods. It has two tiers of complexity. First, allow periods to be associated with geospatial regions. Second, devise a system to represent the fuzziness of these geospatial regions. I am not sure whether both can be tackled in one summer, or if I should pin my focus on one. Thanks for all the input. -Daphne On Sunday, 16 March 2014 05:10:07 UTC-4, Koen Van Daele wrote: Hi Daphne, I did some work on representing vague/imperfect temporal information in relational databases. It's a very interesting, but quite complicated subject. I did a masterpaper on it, but that's in Dutch. I don't suppose you speak Dutch? It does also contain an English abstract. You can find the masterpaper here: http://lib.ugent.be/fulltxt/RUG01/001/418/820/RUG01-001418820_2010_0001_AC.pdf. I did a conference paper on this subject at CAA 2012. You can find the slids here: http://www.slideshare.net
RE: [Arches] Re: Ideas for GSoC Project: Improve Representation of Spatial-Temporal Relationships
Hi Daphne, I think that anything that has to deal with vague temporal or spatial information will automatically turn out te be fairly complex. The fuzzy endpoints of periods is probably the easier one. There the big hurdle is making a good UI that allows for people to enter vague temporal information in a simple and straightforward way that doesn't take to long. We've currently implemented this with a set of widget that allow choosing a start period and an end period whereby both can have different levels of fuzzyness. Eg. A period of time that runs from 'the 8th century BC' till the year 57 BC (Iron Age in Belgium). It's a very textual way of entering data, but that seems to match our users best. Although in actual fact we currently only use the Fuzzy Temporal Intervals for dates of Events (an excavation, a survey, aerial photography, ...) For dating of sites and monuments we use an entirely different approach based solely on a thesaurus of periods. I think the second direction: Allow periods and regions to be associated poses a lot more conceptual issues. I guess one could define a whole set of different Iron Ages, but I have a hard time seeing how it would all work on a practical level. Especially once you start to factor in bringing together data from different implementations and the fact that a region can be vague as well. I also think that another complicating factor can be researches who tend take their chronological system with them. For example, in Flanders, a part of the Iron Age archaeologists use the French division of the Iron Age and another group uses the Dutch division. One group tends to work closer to the French border, the other one closer to the Dutch border. But they also seem to be aligned with the universities the archaeologists graduated from. So I'm not even sure anyone would be capable of geographically defining those ares. Also, when someone then tries to search for Late Iron Age, should they get results from the French Late Iron Age, the Dutch one or both? Or should it be impossible to search for something like 'Late Iron Age' and only possible to search for a specific one? I'm also not sure if the definition of a certain period in a certain area has stayed constant through time. That probably affects things as well. Anyway, it's an absolutely facinating research topic, but one where I think you really need to define a clear scope for your project and resist the tempation of going to far away from it. And make sure that you can communicate to users of your system which choices you've made and why. Cheers, Koen Van: archesproject@googlegroups.com [archesproject@googlegroups.com] namens Daphne Ippolito [daphne...@gmail.com] Verzonden: zondag 16 maart 2014 20:08 Aan: archesproject@googlegroups.com Onderwerp: [Arches] Re: Ideas for GSoC Project: Improve Representation of Spatial-Temporal Relationships Hi. Thanks for your responses. I don't know Dutch, but your abstract did a good job of explaining what the situation is and the terminology I should be looking into. There seems to be two separate directions this project could go in. 1. Improving the way Arches deals with endpoints of periods by allowing for the input of fuzziness. This task doesn't have a spatial component to it. The project would involve implementing a way for users to specify uncertainty and subjectivity in their period definitions using the algorithms you mention in your post. 2. Creating the ability for periods to be spatially referenced. This task only deals with the spatial, not the temporal, definition of periods. It has two tiers of complexity. First, allow periods to be associated with geospatial regions. Second, devise a system to represent the fuzziness of these geospatial regions. I am not sure whether both can be tackled in one summer, or if I should pin my focus on one. Thanks for all the input. -Daphne On Sunday, 16 March 2014 05:10:07 UTC-4, Koen Van Daele wrote: Hi Daphne, I did some work on representing vague/imperfect temporal information in relational databases. It's a very interesting, but quite complicated subject. I did a masterpaper on it, but that's in Dutch. I don't suppose you speak Dutch? It does also contain an English abstract. You can find the masterpaper here: http://lib.ugent.be/fulltxt/RUG01/001/418/820/RUG01-001418820_2010_0001_AC.pdf. I did a conference paper on this subject at CAA 2012. You can find the slids here: http://www.slideshare.net/koenedaele/imperfect-temporal-information-in-data-sets. The digital paper should have been published by now, but the CAA publication process seems to be quite slow at the moment. The basic idea was to treat spans of time as fuzzy sets. I then implemented and tested a few different algorithms on actually reasoning with these fuzzy temporal intervals. The implementation consists of using postgis geometries for representing fuzy
