際際滷

際際滷Share a Scribd company logo

Defining Domain-Specific Facets for Topic Maps With TMQL Path Expressions

T
tmra

The automatic generation of facets works fairly bad for fine-modeled ontologies, in which not all information concerning a single Topic is available through occurrences and direct associations. In this paper, we share our conception of using TMQL path expressions for the definition of domain-specific facets by means of using standard-based Topic Maps technologies. The generated facets must be evaluated, even though they are defined manually by a domain expert. We therefore propose metrics for automatic evaluation of the defined facets, as well as a mechanism for using automatically stored user feedback.

1 of 23
Download to read offline
De鍖ning Domain-Speci鍖c Facets for Topic Maps With TMQL Path Expressions Sven Windisch Dr. Lutz Maicher Topic Maps Lab Natural Language Processing Group Computer Science Department University of Leipzig September 30, 2010 Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 1/23
Agenda Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 2/23
Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 3/23
De鍖nition of facets Facets are . . . . . . clearly de鍖ned, mutually exclusive, and collectively exhaustive aspects, properties or characteristics of a speci鍖c subject. Example Some person has a name, a date of birth and a birthplace. These three characteristics are mutually exclusive and collectively exhaustive. They are facets of this person. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 4/23
Mathematical De鍖nition Any facet can be described as a triple, consisting of the parent identi鍖er id(t) (i. e. whose Topics facet is this) the name of the facet n (e. g. Department, Format, Binding) the value of the facet v (e. g. Paperback, Hardcover ) f = (id(t), n, v) Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 5/23
Famous facet examples Yelp lists lots of different burgers in San Francisco. The facets help you with choosing the one place that 鍖ts your need best. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 6/23
Defining Domain-Specific Facets for Topic Maps With TMQL Path Expressions
Facettable Elements of a Topic Names Names are very useful, but need careful spellchecking. Special Name types (e.g. Forename, Surname) provide a more concise classi鍖cation. Occurrences Occurences with short values are 鍖ne for faceting. Again, careful spellchecking is necessary. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 8/23
Facettable Elements of a Topic Roles A Topic plays a speci鍖c Role, or not. Roles can therefore be used as 鍖ag-like facets. Topic Types Types of Topics are useful for distinguishing different kinds of search results. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 9/23
Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 10/23
Dont they come automagically? Automatic generation of generic facets is possible (Ueberall et al. 2008). But: Information needs of users are highly domain speci鍖c, while generic facets are not. Domain speci鍖c facets must cover more than the the Topics characteristics. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 11/23
Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 12/23
Why TMQL? TMQL makes navigating through a Topic Map very easy But: Full TMQL is much too powerful for our requirements. The path expressions can do the whole job. De鍖nition A path expression represents a sequence of navigation steps through the abstract bidirectional graph of a Topic Map. Starting from given values (atoms or items in a Topic Map), navigation steps along de鍖ned axes within the context map compute new values. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 13/23
How to De鍖ne a Facet Any facet de鍖nition resembles a simple path expression with the following limitations The starting element must be a Topic that serves as Topic type for other Topics. The possible result values are restricted to strings, because the names and values of facets must be human readable. The given Topic type identi鍖er is replaced with the respective identi鍖er of one of the instances. The TMQL expression is then executed and the result is stored as the facets value. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 14/23
Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 15/23
Example Topic Map Mission: Create an of鍖ce-location-facet for every person. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 16/23
Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 17/23
Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 18/23
Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 19/23
Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 20/23
From TMQL to Facets in Three Simple Steps http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name http://psi.example.com/person is replaced by http://psi.example.com/Sven_Windisch The given query is executed, the result is Leipzig A new facet is created: (http://psi.example.com/Sven_Windisch, Of鍖ce Location, Leipzig) Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 21/23
Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 22/23
Conclusions Domain-speci鍖c facets are useful wherever facets must link deeper. TMQL does a wonderful job for de鍖ning domain-speci鍖c facets. De鍖ning domain-speci鍖c facets is easy. The used Topic Map must be well modeled and spellchecked. Outlook Release: Maiana prototype in October, complete search engine at the end of 2010 Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 23/23

More Related Content

Defining Domain-Specific Facets for Topic Maps With TMQL Path Expressions

  • 1. De鍖ning Domain-Speci鍖c Facets for Topic Maps With TMQL Path Expressions Sven Windisch Dr. Lutz Maicher Topic Maps Lab Natural Language Processing Group Computer Science Department University of Leipzig September 30, 2010 Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 1/23
  • 2. Agenda Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 2/23
  • 3. Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 3/23
  • 4. De鍖nition of facets Facets are . . . . . . clearly de鍖ned, mutually exclusive, and collectively exhaustive aspects, properties or characteristics of a speci鍖c subject. Example Some person has a name, a date of birth and a birthplace. These three characteristics are mutually exclusive and collectively exhaustive. They are facets of this person. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 4/23
  • 5. Mathematical De鍖nition Any facet can be described as a triple, consisting of the parent identi鍖er id(t) (i. e. whose Topics facet is this) the name of the facet n (e. g. Department, Format, Binding) the value of the facet v (e. g. Paperback, Hardcover ) f = (id(t), n, v) Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 5/23
  • 6. Famous facet examples Yelp lists lots of different burgers in San Francisco. The facets help you with choosing the one place that 鍖ts your need best. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 6/23
  • 8. Facettable Elements of a Topic Names Names are very useful, but need careful spellchecking. Special Name types (e.g. Forename, Surname) provide a more concise classi鍖cation. Occurrences Occurences with short values are 鍖ne for faceting. Again, careful spellchecking is necessary. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 8/23
  • 9. Facettable Elements of a Topic Roles A Topic plays a speci鍖c Role, or not. Roles can therefore be used as 鍖ag-like facets. Topic Types Types of Topics are useful for distinguishing different kinds of search results. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 9/23
  • 10. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 10/23
  • 11. Dont they come automagically? Automatic generation of generic facets is possible (Ueberall et al. 2008). But: Information needs of users are highly domain speci鍖c, while generic facets are not. Domain speci鍖c facets must cover more than the the Topics characteristics. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 11/23
  • 12. Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 12/23
  • 13. Why TMQL? TMQL makes navigating through a Topic Map very easy But: Full TMQL is much too powerful for our requirements. The path expressions can do the whole job. De鍖nition A path expression represents a sequence of navigation steps through the abstract bidirectional graph of a Topic Map. Starting from given values (atoms or items in a Topic Map), navigation steps along de鍖ned axes within the context map compute new values. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 13/23
  • 14. How to De鍖ne a Facet Any facet de鍖nition resembles a simple path expression with the following limitations The starting element must be a Topic that serves as Topic type for other Topics. The possible result values are restricted to strings, because the names and values of facets must be human readable. The given Topic type identi鍖er is replaced with the respective identi鍖er of one of the instances. The TMQL expression is then executed and the result is stored as the facets value. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 14/23
  • 15. Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 15/23
  • 16. Example Topic Map Mission: Create an of鍖ce-location-facet for every person. Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 16/23
  • 17. Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 17/23
  • 18. Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 18/23
  • 19. Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 19/23
  • 20. Navigating the Topic Map http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 20/23
  • 21. From TMQL to Facets in Three Simple Steps http://psi.example.com/person >> traverse http://psi.example.com/mailing-address >> traverse http://psi.example.com/container-containee-address >> characteristics tm:name http://psi.example.com/person is replaced by http://psi.example.com/Sven_Windisch The given query is executed, the result is Leipzig A new facet is created: (http://psi.example.com/Sven_Windisch, Of鍖ce Location, Leipzig) Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 21/23
  • 22. Overview Introduction to Facets in Topic Maps De鍖ning Domain-Speci鍖c Facets With TMQL A Short Example Conclusions and Outlook Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 22/23
  • 23. Conclusions Domain-speci鍖c facets are useful wherever facets must link deeper. TMQL does a wonderful job for de鍖ning domain-speci鍖c facets. De鍖ning domain-speci鍖c facets is easy. The used Topic Map must be well modeled and spellchecked. Outlook Release: Maiana prototype in October, complete search engine at the end of 2010 Sven Windisch, Dr. Lutz Maicher (Topic Maps Lab) September 30, 2010 23/23