Difference between revisions of "24.07.2013 Trendylyzer"
m (→Outcomes) |
|||
| (8 intermediate revisions by the same user not shown) | |||
| Line 2: | Line 2: | ||
'''Participants''': W. Appeltans (OBIS), M. Flavell (OBIS), A. Italiano (CNR), G. Coro (CNR), A. Ellenbroek (FAO), T. Webb (U. Sheffield) | '''Participants''': W. Appeltans (OBIS), M. Flavell (OBIS), A. Italiano (CNR), G. Coro (CNR), A. Ellenbroek (FAO), T. Webb (U. Sheffield) | ||
| − | |||
| − | |||
| − | |||
== '''Topics''' == | == '''Topics''' == | ||
| + | -Status VRE Trendylyzer developments<br> | ||
| + | -Identifying common species<br> | ||
| + | -Ways forward<br> | ||
| − | + | == '''Outcomes''' == | |
| + | '''VRE'''<br> | ||
| + | G. Coro demonstrated the current development version of the VRE. <br> | ||
| + | The team was positive about progress made so far, and thanked A. Italiano for her good work. Trends on the basis of absolute Nr of records will be one way but we now need to explore additional methods<br> | ||
| + | T. Webb made suggestions to make it possible to select multiple areas, select by taxonomic group and upload a list of species.<br> | ||
| + | A. Ellenbroek suggested that the VRE should be able to e.g.: | ||
| + | <ol> | ||
| + | <li>Identify datasets that hold records of Sarda sarda going back to 1960.</li> | ||
| + | <li>Filter these datasets; only include species that respond to the same sampling technique; e.g. a taxon filter. The idea is to reduce bias. I propose taxon since that seems already available in the infrastructure. Could be at order or class or phylum level. Size would be another. Size of capture is another interesting topic for trendylyzer.</li> | ||
| + | <li>If we have multiple datasets, add option to merge, or keep separate collections. When merging, make sure not to introduce bias by adding many additional species, correct for effort, etc.</li> | ||
| + | <li>Maintain all these filter informations with the result. Even in a graph, there should be a note; "To create this image, x collections were used with a total of y observations. z were used in this analysis"</li> | ||
| + | </ol> | ||
| − | + | '''Common species'''<br> | |
| + | We had some discussions on the definition of common species. Many factors play a role, such as coverage in time, abundance, geographic spread etc<br> | ||
| + | T. Webb and OBIS are currently exploring alternative ways to identify common species<br> | ||
| + | P. Pagano proposed to look at TF-IDF methods to rank species in datasets, but they need to check how much resources are required to develop this method before they decide to commit in this.<br> | ||
| − | + | '''Ways Forward'''<br> | |
| + | OBIS will be asked to evaluate the VRE once it is in production mode.<br> | ||
| + | OBIS and T. Webb will continue testing their methods on selecting common species and see if they produce any meaningful results<br> | ||
| + | If these additional methods prove to be of value the specifications will be shared with the development team for implementation in the VRE<br> | ||
| + | CNR will consider the suggestions from T. Webb and A. Ellenbroek to improve the VRE functionalities.<br> | ||
Latest revision as of 11:59, 25 July 2013
Google Hangout; 24 July 2013 (2-3 PM Brussels Time)
Participants: W. Appeltans (OBIS), M. Flavell (OBIS), A. Italiano (CNR), G. Coro (CNR), A. Ellenbroek (FAO), T. Webb (U. Sheffield)
Topics
-Status VREVirtual Research Environment. Trendylyzer developments
-Identifying common species
Outcomes
VREVirtual Research Environment.
G. Coro demonstrated the current development version of the VREVirtual Research Environment..
The team was positive about progress made so far, and thanked A. Italiano for her good work. Trends on the basis of absolute Nr of records will be one way but we now need to explore additional methods
T. Webb made suggestions to make it possible to select multiple areas, select by taxonomic group and upload a list of species.
A. Ellenbroek suggested that the VREVirtual Research Environment. should be able to e.g.:
- Identify datasets that hold records of Sarda sarda going back to 1960.
- Filter these datasets; only include species that respond to the same sampling technique; e.g. a taxon filter. The idea is to reduce bias. I propose taxon since that seems already available in the infrastructure. Could be at order or class or phylum level. Size would be another. Size of capture is another interesting topic for trendylyzer.
- If we have multiple datasets, add option to merge, or keep separate collections. When merging, make sure not to introduce bias by adding many additional species, correct for effort, etc.
- Maintain all these filter informations with the result. Even in a graph, there should be a note; "To create this image, x collections were used with a total of y observations. z were used in this analysis"
We had some discussions on the definition of common species. Many factors play a role, such as coverage in time, abundance, geographic spread etc
T. Webb and OBIS are currently exploring alternative ways to identify common species
P. Pagano proposed to look at TF-IDF methods to rank species in datasets, but they need to check how much resources are required to develop this method before they decide to commit in this.
Ways Forward
OBIS will be asked to evaluate the VREVirtual Research Environment. once it is in production mode.
OBIS and T. Webb will continue testing their methods on selecting common species and see if they produce any meaningful results
If these additional methods prove to be of value the specifications will be shared with the development team for implementation in the VREVirtual Research Environment.
CNR will consider the suggestions from T. Webb and A. Ellenbroek to improve the VREVirtual Research Environment. functionalities.
