ontoportal-lirmm / annotators

Web service to add functionalities to the http://bioportal.bioontology.org and similar ontology annotators
5 stars 6 forks source link

implement cvalue score #6

Closed julien-diener closed 7 years ago

julien-diener commented 9 years ago

done in 230918e8b6cdafd151e9dcf2ab3b9c7e3319f2fe

jonquet commented 9 years ago

OK...

So testing with the following sentence: cancer du sein cancer du sein cancer du sein cancer du sein cancer cancer with ontology MSHFRE at http://bioportal.lirmm.fr:10000/sifr_annotator using score=cvalue level=0

1 (for Julien). It seems that the scores are not good because in theory... "cancer du sein" should come up in the results.

2 (for Vincent). The JSON link generated at the bottom of the page looks to have an error: http://tubo.lirmm.fr:8080/annotators/sifr_annotator?apikey=1de0a270-29c5-4dda-b043-7c3580628cd5&text=cancer%20du%20sein%20cancer%20du%20sein%20cancer%20du%20sein%20cancer%20du%20sein%20cancer%20cancer%20%0A&max_level=0&ontologies=MSHFRE&score=cvalue

julien-diener commented 9 years ago
  1. in the link you gave, "cancer du sein" is coming up in the results.
  2. but as discussed, the implementation of cvalue is not correct :

    i) annotated terms of concept B that have A in their hierarchy, should be counted in cvalue of A ii) annotated terms should be counted only once in the cvalue part of the score, even if they appear multiple times.

vemonet commented 9 years ago

For the JSON link generated by the GUI : there is no error in the link It is well generated, well encoding "\n" with "%0A" But it seems that the Java Servlet is changing those %0A to \n before calling the annotator.

I have made a little change to the "annotators" project in the develop branch Just adding "replace("\n", "%0A")" when building the URI used to query the annotator : URI uri = new URI(url.replace(" ", "%20").replace("\n", "%0A"));

As you can see on tubo it works fine with "annotators_beta" (which is my war generated from the develop branch) http://tubo.lirmm.fr:8080/annotators_beta/sifr_annotator?apikey=1de0a270-29c5-4dda-b043-7c3580628cd5&text=cancer%0Acancer While it's not working on the actual one http://tubo.lirmm.fr:8080/annotators/sifr_annotator?apikey=1de0a270-29c5-4dda-b043-7c3580628cd5&text=cancer%0Acancer

julien-diener commented 9 years ago

I make another issue #12 for the "\n" problem, not to confuse. And it's resolved.

julien-diener commented 9 years ago

cvalue propagation to hierarchy applies also to mapping

jonquet commented 7 years ago

This has been finished in 2015.