search

King-s-Knowledge-Graph-Lab / ProVe

This tool enables the fact-checking of Wikidata items by verbalizing claims and gathering HTML reference resources utilizing various language models. Please refer to the ProVe API through the link provided below.
https://king-s-knowledge-graph-lab.github.io/ProVe/
4 stars 0 forks source link

Rank / Classify Sources #34

Open salgo60 opened 1 week ago

salgo60 commented 1 week ago

see Wikidata_talk:WikiProject_Reference_Verification

I stated in 2019 that we need to rank sources see T222142 Wikidata has now been used a lot of a research project "Riksdagens Corpus" ( @BobBorges ) and we agree that a sources like Svenskt Biografiskt Lexikon-ID (P3217) / Svenskt biografiskt lexikon (Q379406) / Tvåkammar-riksdagen 1867–1970 (Q110346241) are very good sources, they are just textstrings so to use them in Wikidata its some manual work see issue #78

My suggestion: add a ranking value for sources so more people can agree and understand that e.g. Svenskt Biografiskt Lexikon-ID (P3217) is high quality and have a quality process I think there was some measurement for prizes i.e. that getting the Nobelpriset (Q7191) is ranked higher than getting a prize xxx see my thoughts 2019 that prizes could be a way of evaluating research in different countries... "T216409 Nobelprize as part of evaluating research in different countries"

Maybe we can have dashboards how different research projects support PROV and use quality sources to motivate research to move faster in the right direction....

salgo60 commented 1 week ago

Denny Vrandečić about his vision of sources

BobBorges commented 1 week ago

It would be really good to rank sources if objective criteria could be applied to the ranking.

salgo60 commented 1 week ago

@BobBorges listen to Denny above he tells that en:Wikipedia rank sources. Guess it would be better if the ranking is Done by your project and SBL….

I use Wikidara rank feature and mark wrong facts by e.g. bad precision or not States in the birth record…. —> in the long run we get a rather good quality measurement. I like the way your project test your data against external “sources” like Wikidata but miss that I don’t see SBL in a metadata roundtrip echosystem….