Xper3 (www.xper3.com) is a web platform dedicated to managing descriptive data and to producing interactive identification keys using the webservice Mkey+ (www.identificationkey.fr/mkeyplus). It was first launched at the TDWG 2013 Conference. After only ten months post-launch, Xper3 already counts 485 users and more than 830 databases have been produced.

The Xper3 data model is closely based on the SDD (Structured Descriptive Data) format, which makes it compatible with any other application using this standard format. Due to a Model-View-Controller (MVC) architecture, it is easy to adapt the MKey+ interface for each specific application. For instance, we recently achieved a custom interface for the portal of a citizen science project.

"The Photographic Survey of Flower Visitors” (SPIPOLL, www.spipoll.org, in French) is a French citizen science programme of the Natural history Museum in Paris (MNHN). It aims to obtain quantitative data about pollinating insects and their associated flora in France. Following a simple and entertaining protocol, everyone can photograph insects on flowers, identify the flower using a free access key, and then add pictures and species names in the SPIPOLL database. The data are then validated by experts for scientific use. Because it is used by a large public, the protocol needs an insect identification methodology for neophytes. To this end, the Spipoll team has created a knowledge Xper3 base about pollinating insects. Then, thanks to the implemented effortless export from Xper3 database to SDD format and the interactive identification service (Mkey+), participants can easily identify their own insects. You can find the specific identification service for SPIPOLL at spipoll.snv.jussieu.fr/mkey/mkey-spipoll.html.

Moreover, we set up a mechanism to analyse user behaviours in the identification service. A first development step was to store the identification history (steps of the key) for each picture. The next step will be to investigate these records regarding: user identifications compared to expert identifications, most frequent mistakes, most frequently used characters, etc.  The last step will be to include this feedback in order to compute the most efficient characters in a key context, according to user identification skill and possible species.