Adding OpenTapioca to the available annotators

[wetneb]

Hi!

First, thanks for maintaining GERBIL, it is a very useful tool!
If you are interested, I am running an entity linker for Wikidata at https://opentapioca.org/ which exposes a NIF endpoint at https://opentapioca.org/api/nif. It could be nice to add it to the list of available annotators, although it is already possible to add it manually as a webservice.

[MichaelRoeder]

Added OpenTapioca
@wetneb please check the lines in 93c00d0 whether the name is correct.

However, the benchmark results are currently not correct, e.g., the two results of the annotator are counted as fp and fn although they are overlapping the expected set of URIs.

2019-04-30 12:44:57,628 [pool-1-thread-1] DEBUG [org.aksw.gerbil.matching.impl.MatchingsCounterImpl] - <Found a false negative ((0, 20, [http://dbpedia.org/resource/Florence_May_Harding, http://wikidata.org/entity/Q5460709, http://yago-knowledge.org/resource/Florence_May_Harding, http://en.wikipedia.org/wiki/Florence_May_Harding, http://www.ontologydesignpatterns.org/data/oke-challenge/task-1/Florence_May_Harding, http://wikidata.dbpedia.org/resource/Q5460709, http://rdf.freebase.com/ns/m.0h1d92w])).>
2019-04-30 12:44:57,628 [pool-1-thread-1] DEBUG [org.aksw.gerbil.matching.impl.MatchingsCounterImpl] - <Found a false negative ((44, 6, [http://en.wikipedia.org/wiki/Sydney, http://dbpedia.org/resource/Sydney_(Australia), http://wikidata.org/entity/Q3130, ...])).>
2019-04-30 12:44:57,628 [pool-1-thread-1] DEBUG [org.aksw.gerbil.matching.impl.MatchingsCounterImpl] - <Found a false positive ((44, 6, [http://www.wikidata.org/entity/Q3130])).>
2019-04-30 12:44:57,628 [pool-1-thread-1] DEBUG [org.aksw.gerbil.matching.impl.MatchingsCounterImpl] - <Found a false positive ((0, 20, [http://www.wikidata.org/entity/Q5460709])).>

[wetneb]

Many thanks, that is wonderful! And I am very interested in the scoring issues of course, since it could potentially increase the scores I reported in the paper :)

(By the way it looks like we both used "Tapioca" to name our projects - sorry about this conflict, if I knew I would have picked another name!)

[MichaelRoeder]

The issue is pretty easy 😄
Our sameAs retrieval returns the wrong domain for Wikidata entities, i.e., Sydney is represented as http://wikidata.org/entity/Q5460709 while the annotator returns the correct URI http://www.wikidata.org/entity/Q5460709.

Since http://dbpedia.org/resource/Sydney links to http://www.wikidata.org/entity/Q5460709, I would assume that our index is outdated. @TortugaAttack could you please check whether the issue is caused by outdated data in the index?

(@wetneb no problem. I am happy that you named it "OpenTapioca" which makes it easier to distinguish them 😉 )

[MichaelRoeder]

A run without cached data, without the DBpedia index and with HTTP-based retrieval enabled solves the problem. So the index is simply outdated.
@TortugaAttack we need to update the index.

[wetneb]

If there is anything I can help with to move forward on this, let me know :) From my understanding it is simply a matter of merging 93c00d0 and updating the index, which I cannot do on your instance.

[MichaelRoeder]

Please excuse the long delay. We had some issues to update the indexes.

http://gerbil.aksw.org/gerbil/experiment?id=201908220000
Please close the issue if it works for you 😃

[wetneb]

Thanks a lot! I will add a PR to enable it for A2KB too.