search

isi-vista / cdse-covid

Claim detection & semantic extraction (Covid-19 domain)
0 stars 0 forks source link

Check that events and their arguments are being merged correctly #209

Open elizlee opened 2 years ago

elizlee commented 2 years ago

There are several "duplicate" events between ISI and UIUC in the merged output. I think this may largely be due to the varying offsets identified between each (e.g. UIUC picked out "travelling" while ISI has "travelling undetected"), but I'd like to be sure that's the case and perhaps find a way to handle these.

elizlee commented 1 year ago

The events seem to be getting merged as expected, but their argument statements aren't. If UIUC already has an argument that fills a certain role (e.g. A0), then we don't want to add another one from ISI. An example of this:

<http://www.isi.edu/cdse/assertions/isi/eventarg/L0C049P55/EN_Event_000319/killing.A1/EventArgument-Q43016> a rdf:Statement,
        aida:ArgumentStatement ;
    rdf:object <http://www.isi.edu/gaia/entities/uiuc/L0C049P55/EN_Entity_EDL_ENG_0005022> ;
    rdf:predicate "A1"^^xsd:string ;
    rdf:subject <http://www.isi.edu/gaia/events/uiuc/L0C049P55/EN_Event_000696> ;
    aida:system <http://www.isi.edu/cdse> .
...
<http://www.isi.edu/gaia/assertions/uiuc/eventarg/L0C049P55/EN_Event_000696/Life.Die/Victim/EN_Entity_EDL_ENG_0005022> a rdf:Statement,
        aida:ArgumentStatement ;
    rdf:object <http://www.isi.edu/gaia/entities/uiuc/L0C049P55/EN_Entity_EDL_ENG_0005022> ;
    rdf:predicate "A1_ppt_corpse"^^xsd:string ;
    rdf:subject <http://www.isi.edu/gaia/events/uiuc/L0C049P55/EN_Event_000696> ;
    aida:confidence [ a aida:Confidence ;
            aida:confidenceValue 1e+00 ;
            aida:system <http://www.uiuc.edu> ] ;
    aida:justifiedBy [ a aida:CompoundJustification ;
            aida:confidence [ a aida:Confidence ;
                    aida:confidenceValue 1e+00 ;
                    aida:system <http://www.uiuc.edu> ] ;
            aida:containedJustification <http://www.isi.edu/gaia/assertions/uiuc/eventarg_justification/L0C049P55/EN_Event_000696/Life.Die/Victim/EN_Entity_EDL_ENG_0005022/L0C049P55/4655/4663>,
                <http://www.isi.edu/gaia/assertions/uiuc/eventjustification/EN_Event_000696/L0C049P55/4639/4645> ;
            aida:system <http://www.uiuc.edu> ] ;
    aida:system <http://www.uiuc.edu> .

Source: /nas/gaia/users/elee/phase3_test/cdse_dryrun_0_id_match/WORKING/en/isi_ttl_merged/en/L0C049KSJ.ttl Here we only want to keep the second (UIUC's) A1 argument.