AKTors.org
AKTors.org AKTTechnologiesPublicationsRelated ProjectsPeople
AKTors.org AcquisitionModellingRetrievalReusePublishingMaintenance
NMARKUP from University of Aberdeen


NMARKUP fact-file

Owner  :  University of Aberdeen
Screencam  :  http://www.aktors.org/technologies/nmarkup/nmarkup-3.avi
Builds on  :  GATE - General Architecture for Text Engineering, Java
Addresses challenges  :  Knowledge Modelling

What's the Problem?

Knowledge engineers often build ontologies based on the information given in an appropriate text. That is the knowledge engineer needs to extract pertinent concepts, attributes, values & relations from the text and organize them into a model (ie an ontology). We have noted 2 significant problems with existing tools:

  • A single text may not contain all the entities needed to produce a complete model & hence it is vital to allow the modeler to explore several texts or to allow the Knowledge engineer to add entities from his/her own experience.
  • Most systems only allow users to build a single model/ontology. Whereas a modeler often wishes to develop a family of models & only later select the most plausible one.

Towards a Solution

  • NMARKUP, using the GATE package, processes a text file and highlights all the nouns in the passage.
  • The Knowledge engineer then used NMARKUP's Ontology building subsystem to select concepts, attributes, values and relationships, and to indicate how they are related.
  • The Knowledge Engineer can, if he wish, add entities which he thinks are necessary but which are not found in the text.
  • NMARKUP allows the knowledge engineer create a family of Ontologies
  • These ontologies can be displayed graphically or as RDF; further they can be saved for development in subsequent sessions.

The technology

Below we give a diagram of the NMARKUP system:


The system architecture of NMarkup

Note:

  • GATE produces the information about position of words (POS tags)
  • NMARKUP comprises 2 sub-modules, namely, NounsMarkup and OntoServlet.

We have plans to develop NMARKUP to:

  • Add a mode which allows the modeller to decide to handle all nouns or only unusual nouns (the later are more likely to be domain-specific concepts)
  • Once a word/concept has been extracted from a sentence, bring to the Modeller's attention other unusual words in that same line and adjacent lines;
  • And a variant of the above which allows the user having found an attributes in a line to search the line and adjacent lines for adjectives (ie potential values)
  • Report verbs which might be functioning in the passage as Relations

Try a Demonstration

A video which demonstrates the use of NMARKUP is available (VIDEO)

Further Reading

Key document:

Gang Lei (2003) A Web-Based Tool for Developing Ontologies from Texts. Master of Science (by Research), Computing Science, University of Aberdeen. (This is available on-line on request to Derek Sleeman <dsleeman@csd.abdn.ac.uk>)

Other relevant documents:

H Cunningham, D Maynard, K Bontcheva, V Tablan, C Ursa and M Dimitrov (2003) Developing Language Processing Components with GATE (a User Guide) for GATE version 2.1 (February 2003) http://www.gate.ac.uk/sale/tao/index.html

Semantic representation

View in the AKT Triplestore Browser or as RDF.

Also available in DOAP RDF (Description Of A Project)