Presentation is loading. Please wait.

Presentation is loading. Please wait.

Ontology-enhanced retrieval (and Ontology-enhanced applications) Deborah L. McGuinness Associate Director and Senior Research Scientist Knowledge Systems.

Similar presentations


Presentation on theme: "Ontology-enhanced retrieval (and Ontology-enhanced applications) Deborah L. McGuinness Associate Director and Senior Research Scientist Knowledge Systems."— Presentation transcript:

1 Ontology-enhanced retrieval (and Ontology-enhanced applications) Deborah L. McGuinness Associate Director and Senior Research Scientist Knowledge Systems Laboratory Stanford University Stanford, CA 94305 650-723-9770 dlm@ksl.stanford.edu dlm@ksl.stanford.edudlm@ksl.stanford.edu (FindUR,CLASSIC,PROSE work supported by AT&T Labs Research, Florham Park, NJ, OntoBuilder work supported by VerticalNet, Chimaera, Ontolingua, JTP supported by DARPA)

2 One Conceptual Search u Input is in a natural query language (forms, English, ER diagram …) u Query may be transformed (behind the scenes) into a precise query language with defined semantics u Information is at least semi-structured with DL-like markup and also exists in more natural formats and is interoperable u Answers returned that are not just the explicit answer to question (but also the implicit answer to question) u Answers return the portion of the content that is of use (not an entire page of content) u Answers may be summarized, abstracted, pruned u Answers may be services that can take action u Interface is interactive and helps users reformulate unsuccessful queries u Customizable, extensible, …

3 Today: Rich Information Source for Human Manipulation/Interpretation Human

4 I know what was input u Global documents and terms indexed and available for search u Search engine interfaces u Entire documents retrieved according to relevance (instead of answers) u Human input, review, assimilation, integration, action, etc. u Special purpose interfaces required for user friendly applications The web knows what was input but does little interpretation, manipulation, integration, and action

5 Information Discovery… but not much more u Human intensive (requiring input reformulation and interpretation) u Display intensive (requiring filtering) u Not interoperable u Not agent-operational u Not adaptive u Limited context u Limited service Analogous to a new assistant who is thorough yet lacks common sense, context, and adaptability

6 Future: Rich Information Source for Agent Manipulation/Interpretation HumanAgent

7 I know what was meant u Understand term meaning and user background u Interoperable (can translate between applications) u Programmable (thus agent operational) u Explainable (thus maintains context and can adapt) u Capable of filtering (thus limiting display and human intervention requirements) u Capable of executing services

8 One Approach… start simple from embedded bases u Recognize the vast amount of information in textual forms… u Enhance standard information retrieval by adding some semantics u Use background ontology to do query expansion u Exploit ontology to add some structure to IR search u Move to parametric search u Move to include inference (in e-commerce setting moving towards interoperable solutions and configuration

9 FindUR Challenges/Benefits FindUR Challenges/Benefits u Retrieve documents otherwise missed - Recall u More appropriately organize documents according to relevance (useful for large number of retrievals) u Browsing support (navigation, highlighting) u Simple User Query building and refinement u Full Query Logging and Trace u Facilitate use of advanced search functions without requiring knowledge of a search language u Automatically search the right knowledge sources according to information about the context of the query

10 ( FindUR Architecture Search Engine Content to Search: Search and Representation Technology: User Interface: Verity Topic Sets Content (Web Pages, Documents, Databases) Results (domain spec.) Verity SearchScript, Javascript, HTML, CGI Content Classification Domain Knowledge Results (std. format) Search Parameters Classic Collaborative Topic Building Tool Query Input P-CHIP Research Site Technical Memorandum Calendars (Summit 2005, Research) Yellow Pages (Directory Westfield) Newspapers (Leader) AT&T Solutions Worldnet Customer Care

11

12

13

14

15 OntologyBuilder

16 Configuration http://www.research.att.com/sw/tools/classic/tm/ijcai-95-with-scenario.html

17 Ontology Creation and Maintenance Environment Needs u Semi-automatic generation input u Diagnostics/Explanation (Chimaera, CLASSIC,…) u Merging and Difference (Chimaera, Prompt, Ontolingua, …) u Translators/Dumping (Ontolingua, …) u Distributed Multi-User Collaboration (OntologyBuilder,…) u Versioning (OntologyBuilder,…) u Scalability. Reliability, Performance, Availability (Shoe,OntologyBuilder,…) u Security (viewing, updates, abstraction, authoritative sources…) u Ontology Library systems (Ontolingua,…) u Business needs – internationalization, compatibility with standards (XML,…)

18 Conclusion With background ontologies and the appropriate environments, we can move from simple ontology-enhanced applications to the next generation web

19 Pointers u FindUR: www.research.att.com/~dlm/findurwww.research.att.com/~dlm/findur u OntoBuilder/OntoServer: http://www.ksl.stanford.edu/people/dlm/papers/ontologyBui lderVerticalNet-abstract.html u Deborah McGuinness: www.ksl.stanford.edu/people/dlmwww.ksl.stanford.edu/people/dlm u CLASSIC: www.research.att.com/sw/tools/classicwww.research.att.com/sw/tools/classic u Chimaera: www.ksl.stanford.edu/software/chimaera/


Download ppt "Ontology-enhanced retrieval (and Ontology-enhanced applications) Deborah L. McGuinness Associate Director and Senior Research Scientist Knowledge Systems."

Similar presentations


Ads by Google