We have collected all generated alignments and make them available in a zip-file via the following link. These alignments are the raw results that the following report is based on.
>>> download raw results
We conducted experiments by executing each system in its standard setting and we compare precision, recall, F-measure and recall+. The measure recall+ indicates the amount of detected non-trivial correspondences. The matched entities in a non-trivial correspondence do not have the same normalized label. The approach that generates only trivial correspondences is depicted as baseline StringEquiv in the following section. We used a different machine compared to previous years. Thus, runtime results are not fully comparable across years.
Same as the last year, we run the systems on a server with 3.46 GHz (6 cores) and 8GB RAM allocated to the matching systems. Further, we used the SEALS
client to execute our evaluation. However, we slightly changed the way how precision and recall are computed, i.e., the results generated by the SEALS client vary in some cases by 0.5% compared to the results presented below. In particular, we removed trivial correspondences in the oboInOwlnamespace like
http://...oboInOwl#Synonym = http://...oboInOwl#Synonym
as well as correspondences expressing relations different from equivalence. Using the Pellet reasoner we also checked whether the generated alignment is coherent, i.e., there are no unsatisfiable concepts when the ontologies are merged with the alignment.
In the following, we analyze all participating systems that could generate an alignment in less than ten hours. The listing comprises of 10 entries. There were 2 systems which participated with different versions. These are AOT with versions AOT and AOTL, LogMap with four different versions LogMap, LogMap-Bio, LogMap-C and a lightweight version LogMapLite that uses only some core components. In addition to LogMap and LogMapLite we have 3 more systems which participated in 2013 and now participated with new versions (AML, MaasMatch, XMap). For more details, we refer the reader to the papers presenting the systems. Thus, 10 different systems generated an alignment within the given time frame. There were four participants (InsMT, InsMTL, OMReasoner and RiMOM-IM) that threw an exception or produced an empty alignment and are not considered in the evaluation.
We have 6 systems which finished in less than 100 seconds, compared to 10 systems in OAEI 2013 and 8 systems in OAEI 2012. This year we have 10 out of 13 systems which generated results which is comparable to last year when 20 out of 24 systems generated results within the given time frame. The top systems in terms of runtimes are LogMap, XMap and AML. Depending on the specific version of the systems, they require between 5 and 30 seconds to match the ontologies. The table shows that there is no correlation between quality of the generated alignment in terms of precision and recall and required runtime. This result has also been observed in previous OAEI campaigns.
The table also shows the results for precision, recall and F-measure. In terms of F-measure, the top ranked systems are AML, LogMap-Bio, LogMap and XMap. The latter two generate similar alignments. The results of these four systems are at least as good as the results of the best systems in OAEI 2007-2010. AML has the highest F-measure up to now. Other systems in earlier years that obtained an F-measure that is at least as good as the fourth system this year are AgreementMaker (predecessor of AML) (2011, F-measure: 0.917), GOMMA-bk (2012/2013, F-measure: 0.923/0.923), YAM++ (2012/2013, F-measure 0.898/0.905), and CODI (2012, F-measure: 0.891).This year we have 7 out of 10 systems which achieved an F-measure that is higher than the baseline which is based on (normalized) string equivalence (StringEquiv in the table). This is a better result (percentage-wise) than the last year but still lower than in OAEI 2012 when 13 out of 17 systems produced alignments with F-measure higher than the baseline. Both systems, XMap and MaasMatch, which participated in the last year and had results below the baseline, achieved better results than the baseline this year.
Moreover, nearly all systems find many non-trivial correspondences. Exception are systems RSDLWB and AOTL that generate an alignment that is quite similar to the alignment generated by the baseline approach.
There are 5 systems which participated in the last year, AML, LogMap, LogMapLite, MaasMatch and XMap. From these systems LogMap and LogMapLite achieved the identical results as the last year, while AML, MassMatch and XMap improved their results. MaasMatch and XMap showed a considerable improvement. In the case of MaasMatch, its precision was improved from 0.359 to 0.914 (and the F-measure from 0.409 to 0.803) while XMap which participated with two versions in the last year increased its precision from 0.856 to 0.94 (and F-measure from 0.753 to 0.893) compared to the XMapSig version which achieved a better F-measure last year.
A positive trend can be seen when it comes to coherence of alignments. Last year only 3 systems out of 20 produced a coherent alignment while this year half of the systems produced such alignment.
This year 14 systems participated in the anatomy track out of which 10 produced results. This is a significant decrease in the number of participating systems. However, the majority of the systems which participated in the last year significantly improved their results.
As last year, we have witnessed a positive trend in runtimes as all the systems which produced an alignment finished execution in less than an hour. Same as the last year, the AML system set the top result for the anatomy track by improving the result from the last year. The AML system improved in terms of all measured metrics.
We would like to thank Christian Meilicke for his advices and support with the organization of this track.
This track is organized by Zlatan Dragisic, Valentina Ivanova and Patrick Lambrix. If you have any problems working with the ontologies, any questions related to tool wrapping, or any suggestions related to the anatomy track, feel free to write an email to oaei-anatomy [at] ida [.] liu [.] se.