Hatana - Virtual Topic Map Merging

TMRA 2010

Hatana A virtual merging engine

Uta Schulze Topic Maps Lab at the University of Leipzig

[email protected]‐leipzig.de

Einführung in Topic Maps

Uta Schulze, Topic Maps Lab

<[email protected]>

"Accessing the island of Hatana is a complicated process."

http://en.wikipedia.org/wiki/Hatana



<[email protected]>

MoBvaBon -‐ Not that complicated

Several data sources

One view?



<[email protected]>



One query language?



<[email protected]>



?



<[email protected]>

Why puJng data into topic maps?

Because we can.

MaJorToM-JLI

ARNotations

Coming soon

Maiana



<[email protected]>


Because we can.

Because information wants to be a topic map.



<[email protected]>


Because we can.

Because information wants to be a topic map.

Because we’d like to merge…



<[email protected]>

Why don‘t we just put everything into one big topic map and are done with?

Information

1.  changes over time



<[email protected]>


Information


2.  has copyright issues



<[email protected]>


Information



3.  has an origin that would get lost



<[email protected]>


Information



3.  has an origin that would get lost

4.  should sometimes remain in its database



<[email protected]>

Another soluBon: Hatana

Hatana creates a layer over several data sources.



<[email protected]>

Another soluBon: Hatana

This layer behaves as a topic map!

Hatana creates a layer over several data sources.



<[email protected]>

Example 1: TMRA ParBcipants (Java Live IntegraBon)



<[email protected]>

Hatana merges ...

•  strictly according to the TMDM’s equality rules of constructs

•  on demand

•  and creates „virtual” topics, associations, …



<[email protected]>

Equality rules for topics

Graham Moore,

Vice Admiral

Graham Moore,

Footballer

Graham Moore,

Topic Mapper



<[email protected]>

Graham Moore,

Vice Admiral

Graham Moore,

Footballer

Graham Moore,

Topic Mapper




<[email protected]>

Graham Moore,

Vice Admiral

Graham Moore,

Footballer

NAME EQUALITY

Graham Moore,

Topic Mapper




<[email protected]>

Graham Moore,

Vice Admiral

Graham Moore,

Footballer

Graham Moore,

Topic Mapper

http://en.wikipedia.org/wiki/Graham_Moore

http://www.topicmapslab.de/people/Graham_Moore

http://en.wikipedia.org/wiki/Graham_Moore_(footballer)




<[email protected]>

Graham Moore,

Vice Admiral

Graham Moore,

Footballer

Graham Moore,

Topic Mapper

http://en.wikipedia.org/wiki/Graham_Moore


http://en.wikipedia.org/wiki/Graham_Moore_(footballer)


IDENTIFIER EQUALITY



<[email protected]>

Virtual on demand merging




<[email protected]>





<[email protected]>





<[email protected]>



http://www.topicmapslab.de/people/Graham_Moore http://psi.ontopedia.net/Graham_Moore



<[email protected]>






<[email protected]>






<[email protected]>




http://psi.ontopedia.net/Graham_Moore



<[email protected]>




http://psi.ontopedia.net/Graham_Moore



<[email protected]>



virtual topic



<[email protected]>


virtual topic

get all names

virtual names



<[email protected]>

Virtual topic

•  empty

•  wrapper that

•  acts as topic ...

•  but stores ids of equal source topics




<[email protected]>

Example 2: Merging the Italian and Norwegian Opera

Problem

•  only 20 out of 2767 topics with equal identifiers

•  http://psi.ontopia.net/music/opera vs. http://psi.ontopedia.net/Opera

•  no editing of sources



<[email protected]>

Example 2: Merging the Italian and Norwegian Opera

Problem

•  Only 20 out of 2767 topics with same identifiers

•  http://psi.ontopia.net/music/opera vs. http://psi.ontopedia.net/Opera

•  No editing of sources

Possible solution

•  topic map with topic containing e.g. both „Opera“ identifiers

•  source: expert knowledge, Subj3ct, ...



<[email protected]>

Advantages of virtual merging

•  Combining read-only topic maps

•  Combining private with public available topic maps (Maiana feature)

•  Information about the origin

•  On demand merging

•  Playing around with identities

•  Validating topic map against a schema, database backend



<[email protected]>

Performance

•  There is runtime

•  Highly dependent of the underlying engine



<[email protected]>

Performance



•  Validation of the opera map against the TMCL Meta Schema:

•  hard merge using Ontopia: instantly

•  improved Hatana version: took quite some time



<[email protected]>

Performance





•  more improved Hatana version: a moment



<[email protected]>

Performance





•  future Hatana version: ?



<[email protected]>

ImplementaBon

•  read-only Java topic maps engine (TMAPI)

•  Topic Map System containing virtual topic maps

•  Virtual construct: empty layer storing the ids of its source constructs

•  Caching: Id-Storing



<[email protected]>

I did not talk about

•  Merging topics that reify e.g. equal names

•  Merging associations and roles

•  most time consuming

•  Cache invalidation



<[email protected]>

Next steps

•  Learning from Jack Park’s merge assertions

•  Implementing the Container in Container feature

•  Increasing performance

•  Improving the Container view in Maiana



<[email protected]>

Thank you for your attention!

Hatana - Virtual Topic Map Merging

Education

topic maps lab

moore uta schulze

moore http

orgwikihatana uta schulze

lostuta schulze

timeuta schulze

databaseuta schulze

orgw moore