scieee Science in your language
[en] (orig)

User-oriented exploration of semi-structured datasets

Author: Barret, Nelly
Publisher: Zenodo
DOI: 10.5281/zenodo.17679385
Source: https://zenodo.org/records/17679385/files/lisn-slides.pdf
Use -o ien ed explo a ion o semi-s uc u ed da ase s
Nelly Ba e
3 d yea PhD s uden
Supe ised by Ioana Manolescu
In ia Saclay and Ins i u Poly echnique de Pa is
Oc obe 9, 2023
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 1 / 38
Con ex : da a is he new gold (1/3)
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 2 / 38
Con ex : da a is he new gold (1/3)
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 3 / 38
Con ex : da a is he new gold (2/3)
Ou digi al wo ld comes:
In a ious con ex s: science, heal h, poli ical li e
A a ious scales: home, ci y, coun y, wo ld
By di e en ac o s: scien is s, businesses, policy make s
Wi h di e en needs, cons ain s, abili ies
We a e o e whelmed by ( aw) da a, we need:
Da a-d i en applica ions
Da a jou nalism
Knowledge g aphs
A i icial “in elligence”
...
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 4 / 38
Con ex : da a is he new gold (3/3)
Ve y he e ogeneous da a:
Mainly RDF (1K da ase s in he LODC)
Also: XML, JSON, ela ional, P ope y
G aph...
De ec ion o en i ies o in e es :
People, Place, email, ...
Wi h he e ogeneous da a, use s need:
1Auni o m in eg a ion, iew
2E icien algo i hms and applica ions
3A global unde s anding,desc ip ion
4In e es ing en i y connec ions
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 5 / 38

C ea e a unique da a g aph
“A uni o m in eg a ion, iew”
Angelos Anadio is Oana Balalau Ioana Manolescu e al...
IPP, EPFL In ia, IPP In ia, IPP INSEC, ...
He e ogeneous
inpu Loading Collec ion
g aph Abs a
Pa hWays
New
ea u es
S udio
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 6 / 38
G aph cons uc ion
Inges any da ase in o a di ec ed g aph (•,→)
Ex ac named en i ies, NEs, om he g aph alues (◦,99K):
Tempo al: da e , ime e e ence
Web: URI, email add ess , hash ag, Twi e ci a ion
Complex en i ies: People , Place, O ganiza ion
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 7 / 38
C ea e a compac ep esen a ion o he da a g aph
“E icien algo i hms and applica ions”
He e ogeneous
inpu Loading Collec ion
g aph Abs a
Pa hWays
New
ea u es
S udio
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 8 / 38
A uni o m iew o da a o ma s
Each da a o ma has i s own speci ici ies:
RDF (URIs) JSON (labels) XML (ID-IDREF) PG (edge a s)
Bu , we encode he same logic:
Reco d: piece o da a, an objec
Value: eco d wi h no child en
Same-kind eco ds: schema o in ui i e o de
Rela ionship: how eco ds ela e
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 9 / 38
Collec ions weigh s, bounda ies and g aph upda es
Collec ion weigh
wdesck
wlea k
wDAG
wPageRank
wdwPageRank
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 15 / 38

Collec ions weigh s, bounda ies and g aph upda es
Collec ion weigh
wdesck
wlea k
wDAG
wPageRank
wdwPageRank
Bounda y
bounddesc
boundlea
boundDAG
bound lood
boundacyclic− lood
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 16 / 38
Collec ions weigh s, bounda ies and g aph upda es
Collec ion weigh
wdesck
wlea k
wDAG
wPageRank
wdwPageRank
Bounda y
bounddesc
boundlea
boundDAG
bound lood
boundacyclic− lood
G aph upda e
upda eboolean
upda eexac
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 17 / 38
Find ela ionships be ween main collec ions
Possible ela ionships
The se o ela ionships connec ing a pai o collec ions is he se o hei
pa hs.
pape →wB →au ho
pape →pIn →con
au ho →hW →pape
con →in →au ho
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 18 / 38
The inal ou pu in Abs a
h ps:// eam.in ia. /ceda /p ojec s/abs a/
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 19 / 38
Enume a e en i y pa hs
“In e es ing en i y connec ions”
Nelly Ba e An oine Gauquie Jia Jean Law Ioana Manolescu
In ia, IPP IMT IPP In ia, IPP
He e ogeneous
inpu Loading Collec ion
g aph Abs a
Pa hWays
New
ea u es
S udio
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 20 / 38

Pa hWays: ind in e es ing connec ions in he da a
P oblem s a emen
How o in e ac i ely explo e en i y connec ions in he e ogeneous da ase s?
1No que y w i ing, no p io knowledge
2A abula , high-le el ou pu , easy o g asp o NTUs
3Do i e icien ly e en i he da a g aph is la ge
=⇒Connec named en i ies (People, Places, ...) in and ac oss da ase s.
Keywo d G aph Reachabili y Pa hWays
sea ch que y que y
No que y w i ing X× × X
Tabula ou pu ∼ ∼ ×X
E icien ×X X X
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 21 / 38
Pa hWays: ind in e es ing connec ions in he da a
P oblem s a emen
How o in e ac i ely explo e en i y connec ions in he e ogeneous da ase s?
1No que y w i ing, no p io knowledge
2A abula , high-le el ou pu , easy o g asp o NTUs
3Do i e icien ly e en i he da a g aph is la ge
=⇒Connec named en i ies (People, Places, ...) in and ac oss da ase s.
Keywo d G aph Reachabili y Pa hWays
sea ch que y que y
No que y w i ing X× × X
Tabula ou pu ∼ ∼ ×X
E icien ×X X X
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 21 / 38
Scena io and e minology
Ada a (en i y) pa h is a pa h in he da a g aph
Acollec ion (en i y) pa h is a pa h in he collec ion g aph
The e alua ion o a collec ion pa h leads o a se o da a pa hs
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 22 / 38
Collec ion (en i y) pa h enume a ion
(τ1, τ2); max pa h leng h; non-speci ic connec ions
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 23 / 38
Quick o e iew o expe imen s
On widely-used open da a o ma s: JSON, RDF, XML and PG.
Abs a Pa hWays
•Use s udy •# pa hs: 0 o e y high
•Compa ison o schemas •4 pa h “shapes”
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 30 / 38

Fu u e wo k, akeaways and open ques ions
He e ogeneous
inpu Loading Collec ion
g aph Abs a
Pa hWays
New
ea u es
S udio
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 31 / 38
Fu u e wo k: S udio
S udio: a da a lake o inges ing, que ying, cleaning and unde s anding
he e ogeneous da a
F ench media a e in e es ed (Da aJou nos, CFI)
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 32 / 38
Fu u e wo k: S udio
S udio: a da a lake o inges ing, que ying, cleaning and unde s anding
he e ogeneous da a
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 33 / 38
Fu u e wo k: S udio
S udio: a da a lake o inges ing, que ying, cleaning and unde s anding
he e ogeneous da a
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 34 / 38
Fu u e wo k: S udio
S udio: a da a lake o inges ing, que ying, cleaning and unde s anding
he e ogeneous da a
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 35 / 38

Fu u e wo k: S udio
S udio: a da a lake o inges ing, que ying, cleaning and unde s anding
he e ogeneous da a
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 36 / 38
Takeaways and open ques ions
Abs a: a da ase abs ac ion sys em o he e ogeneous da a
Pa hWays: an en i y- ocused explo a ion sys em
S udio: a use -o ien ed da a lake o da a explo a ion
Abs a Pa hWays S udio
EDBT 2024 ADBIS 2023 CoopIS 2023
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 37 / 38
Fu he oppo uni ies
Nelly BARRET
¥[email p o ec ed]
§h ps://pages.saclay.in ia. /nelly.ba e /
In ia Saclay & Ins i u Poly echnique de Pa is
Palaiseau
Nelly Ba e (In ia) Semi-s uc u ed Da a Explo a ion Oc obe 9, 2023 38 / 38