Taxonomies,  Ontologies  and  Knowledge  Graphs  

Faceted  Taxonomies:  Start  Search  From  Any  Perspec>ve,  Progressively  Filter  Results  

Taxonomy    •  List  or  Hierarchy  •  BT/NT/RT  rela>onships  •  Matrix  and  Facet  forms  •  Visual  naviga>on  •  Human  friendly  •  Built  for  known  needs    

Thesaurus    •  BT/NT/RT/  USE/UF  rela>onships  

•  Dic>onary  naviga>on    

Ontology    •  Maps  unlimited  types  of  rela>onship  between  

concepts  •  Difficult  to  build  and  maintain  -­‐    especially  for  

fluid,  imprecise  environments  •  Machine  friendly  not  human  friendly  •  Built  for  mul>ple  poten>al  needs    

Taxonomy    

•  List  or  Hierarchy  •  BT/NT/RT  rela>onships  •  Matrix  and  Facet  forms  •  Visual  naviga>on  •  Human  friendly  •  Built  for  known  needs  

Breathe….  Now  we’re  going  to  talk  about  Knowledge  Graphs,  Taskonomies  and  

leading  into  Search  Based  Apps  Ontology  

Taxonomy  Facets  

Salient  Rela7onships  

Contextual  Analysis  

Content  Sources  

•  Between  facets  •  Between  concepts  •  Between  facets  and  

content  •  Between  content  

and  content  •  Mapping  is  ac>ve  

and  passive  

•  People  •  Places  •  Ins>tu>ons  

•  Wikipedia  •  Weather  server  •  Time  server  •  Map  server  •  Reviews  •  Web  pages  •  Structured  

databases  

•  Taskonomy  –  user  needs  •  Disambigua>on  cards  •  Frequent  next  ac>on  

Relevance  vs  Salience  

Relevance    Retrieved  results  meet  the  need  expressed  in  the  query.    Retrospec>ve,  against  defined  needs.    Supports  finding  stuff  you  are  looking  for.    

Salience    Having  the  quality  of  standing  out  as  especially  significant.    Has  con>nuing  u>lity,  meets  as-­‐yet  unexpressed  needs.    Supports  discovery  of  stuff  you  didn’t  know  existed.    

Taxonomy  vs  Ontology  vs  Knowledge  Graph  

Taxonomy    •  BT/NT/RT  rela>onships  •  Hierarchical  •  Visual  naviga>on  •  Human  friendly  •  Built  for  known  need    Thesaurus    •  BT/NT/RT/  USE/UF  

rela>onships  •  Dic>onary  naviga>on    

Ontology    •  Maps  unlimited  types  of  

rela>onship  between  concepts  

•  Difficult  to  build  and  maintain  -­‐    especially  for  fluid,  imprecise  environments  

•  Machine  friendly  not  human  friendly  

•  Built  for  mul>ple  poten>al  needs  

 

Knowledge  Graph    •  Maps  salient  en>>es  and  

rela>onships  •  Can  map  to  content  as  

well  as  concepts  •  Highly  scalable,  flexible,  

evidence-­‐based,  built  for  known  need  

 

Taskonomy:  an  arrangement  to  enable  common  work  tasks  

“I’m  prescribing  medica>on  for  a  pa>ent.  I  want  to  immediately    iden>fy  medica>on  appropriate  to  a  diagnosis,  that  we  have  in  stock.  I  want  to  check  contraindica>ons  against  my  pa>ent’s  medical  history.  I  want  to  be  aware  of  any  incidents  involving  the  selected  medica>on,  and  I  want  to  be  able  to  advise  the  pa>ent  on  how  to  manage  their  daily  life  while  on  the  medica>on.”  

Search  Based  Applica>ons  are  fed  by  Taskonomies  (+  Taxonomies  +  Knowledge  Graphs)  

Task  

Pharmaceu>cals  database  

(subscrip>on)  

ePharmacy  (hospital)  

Pa>entslikeus.com  (social)  

Pa>ent  records  system  (hospital)  

Incident  repor>ng  system  (hospital)  

“I’m  prescribing  medica>on  for  a  pa>ent.  I  want  to  immediately    iden>fy  medica>on  appropriate  to  a  diagnosis,  that  we  have  in  stock.  I  want  to  check  contraindica>ons  against  my  pa>ent’s  medical  history.  I  want  to  be  aware  of  any  incidents  involving  the  selected  medica>on,  and  I  want  to  be  able  to  advise  the  pa>ent  on  how  to  manage  their  daily  life  while  on  the  medica>on.”  

Pharmaceu>cals  database  

(subscrip>on)  

ePharmacy  (hospital)  

Pa>entslikeus.com  (social)  

Pa>ent  records  system  (hospital)  

Incident  repor>ng  system  (hospital)  

“I’m  prescribing  medica>on  for  a  pa>ent.  I  want  to  immediately    iden>fy  medica>on  appropriate  to  a  diagnosis,  that  we  have  in  stock.  I  want  to  check  contraindica>ons  against  my  pa>ent’s  medical  history.  I  want  to  be  aware  of  any  incidents  involving  the  selected  medica>on,  and  I  want  to  be  able  to  advise  the  pa>ent  on  how  to  manage  their  daily  life  while  on  the  medica>on.”  

Taxonomies  &  Thesauri    •  Resolve  vocabularies  to  common  terms  

•  Support  naviga>on  

Knowledge  graphs  •  Map  rela>onships  

between  concepts  &  concepts  and  between  concepts  and  content  

•  That  are  salient  to  the  task  at  hand