SLIDE 1
Trees of knowledge
Encyclopédie
Commerce, see also Rhetoric: cross-discipline relationships as - - PowerPoint PPT Presentation
Commerce, see also Rhetoric: cross-discipline relationships as - - PowerPoint PPT Presentation
UDC Seminar 2015, Lisbon Commerce, see also Rhetoric: cross-discipline relationships as authority data for enhanced retrieval Claudio Gnoli (U Pavia) Rodrigo De Santis (IFPR) Laura Pusterla (U Pavia) Trees of knowledge Encyclopdie Hugh
SLIDE 2
SLIDE 3
Hugh of Saint Victor (1096?-1141)
commerce could be classified among the practical arts, but also under rhetoric, as "a peculiar sort of rhetoric [...] for eloquence is in the highest degree necessary to it" [in Olson 2010]
SLIDE 4
Commerce, see also...
A concept can occur under several branches...
(WebDewey)
“See also” references in classification schedules
SLIDE 5
Search expansion
“ a systematic (classification) display can be adapted for search expansion, enhancing both recall and precision, through exploiting hierarchical and associative relationships, provided that semantic linking in the vocabulary is fully supported by vocabulary data ” [Slavic 2006]
SLIDE 6
QR codes on shelves?
SLIDE 7
Context provided on shelves
Social Sciences Library, University of Florence links between shelves and the catalogue [Fabbrizzi 2014]
SLIDE 8
Science and Technology Library, U Pavia
SLIDE 9
SLIDE 10
SLIDE 11
Cross-discipline relationships at STL
690 building ← 624 civil engineering 711.4 city planning ← 307 communities 720 architecture ← 690 building 725.5 health buildings ← 610.28 biomedical instrumentation
SLIDE 12
Query expansion
SLIDE 13
SLIDE 14
Relationships as authority data
SLIDE 15
Existential dependence
Dependence in ontology (Lowe 2015) Architecture dependsOn Building dependsOn CivilEngineering
between disciplines (DDC)
between phenomena (ILC)
building architecture civil engineering
SLIDE 16
Relationships as linked data
<owl:Class rdf:about="http://www.iskoi.org/ilc/ilc_ontology#vc"> <rdfs:label>vc</rdfs:label> <rdfs:comment xml:lang="en">fishing</rdfs:comment> <rdfs:subClassOf rdf:resource="http://www.iskoi.org/ilc/ ilc_ontology#v"/> <owl:Restriction> <owl:onProperty rdf:resource="http://www.iskoi.org/ ilc/ilc_ontology#dependsOn" /> <owl:allValuesFrom rdf:resource="http://www.iskoi.
- rg/ilc/ilc_ontology#mq" />
</owl:Restriction> </owl:Class>
SLIDE 17
Relationships as linked data
Forestry dependsOn Plants
SLIDE 18
Problem: too many results
Mur Murph phy's L y's Law of aw of I IR: R: as recall increases, precision decreases
SLIDE 19
Strategies to limit overload
Make expanded search optional
- ptional only
So Sort results e.g. by decreasing publication date
Allow for query refine efinemen ment
Carefully sele select ct cross-discipline relationships:
not always reciprocal
- nly one step at a time [Tudhope et al. 2001]
720 ← 690, 624 690 ← 624
SLIDE 20
Tuning SciGator
Advantage: we can add / edit classes, captions, and relationships according to practical experience with our collection Only classmarks with ≥ 5 owned books are shown (Non official DDC captions) Feedback between reference service and catalogue
SLIDE 21
Integrating collections through classification
DDC as a common language to gather data from different libraries into one catalogue and discovery tool
SLIDE 22
Mapping local schemes
SLIDE 23
Discussion
KOSs imply complex graphs of relationships
hierarchical relationships are powerful, but they are not enough
existential dependence is another key relationship
both planes are needed:
conceptual (identifying relevant relationships)
technical (representing them as authority data)
knowledge discovery tools can be built on these bases
SLIDE 24