- From: Christophe Dupriez <christophe.dupriez@destin.be>
- Date: Tue, 19 Jul 2011 20:30:02 +0200
- To: 'Skos' <public-esw-thes@w3.org>
To those broadly interested to the narrow issues of hierarchies: Looking at how ASKOSI.org performs with the MeSH, I looked to ingestion error messages and found that there was an apparent loop between descriptor "Morals" and descriptor "Ethics". http://www.nlm.nih.gov/mesh/2011/mesh_browser/MBrowser.html (type "Morals" in the search box) Reporting this to NLM, I was kindly recommended to read: http://www.nlm.nih.gov/mesh/meshrels.html where one can find (section 6): "Since its hierarchical relationships are between descriptors, there are practical reasons for a MeSH descriptor to have different children in different trees." At Belgium Poison Centre, we may have to classify substances differently when considering: 1) usage, 2) molecule geometry (chemistry), 3) toxicological impact. Hierarchies (narrower/broader relations) would therefore pertain to one or more "classification methodologies" (facet?) (with potentially different "siblingKey" to sort narrower concepts within each classification methodology) I remain unconvinced that morals and ethics should loop (and that one can allow a broader/narrower relation to reverse with different classification methodologies) but the idea of different hierarchical contexts seems good to me. I remained also unsure a "facet" would allow to represent adequately the MeSH hierarchies. So I fiddled a lot but I was unable to find a more orthogonal representation of MeSH hierarchies than the "Tree Numbers" ( MRREL in UMLS is not sufficient to rebuild them ). * Tree Numbers have the default to need to be recomputed when new concepts are inserted. * They have the advantage of fine grain control over hierarchies and display order. Do you have experience with this kind of issues? Wishing you a very nice evening! Christophe
Received on Tuesday, 19 July 2011 19:09:27 UTC