ALERT! Under construction ALERT!

Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction Work in progress, under construction

Term sense alignment in parallel ontologies using syntactic corpus analysis

Seppo Nyrkkö

University of Helsinki, Department of Modern Languages

Abstract

Background

Aligning ontologies has been a The failing causes for character-similarity alignments are various: Authors name entities differently due to linguistic, cultural and dialectical differences. There is also a notable difference in structures, due to different applications and aims of use.

Approaches

First, we took 3 ontologies:

TAP[-], SUMO[-] and PuLS? [-]

MeSH?

Biocaster?

How syntactic dependencies are used.

*Extracting knowledge in lexicons, dictionaries, ontologies*

Workflow model

patt.txt -->  "dep2prolog"

( )
|_| parsed corpus

  graph matching --> results

Sample of results

s-id tokenA tokenB rule w
42 malaria parasites genus plasmodium memberOf 0.8
42 malaria parasites plasmodium memberOf 0.6
42 parasites genus plasmodium memberOf 0.4

Alignment model

Aligning syntactic dependencies with ontological relations

Token alginment

Predicate alignment

Methods

Similarity / distance model

  • Hop models
    • Synonymity
    • Bilingual pairs

Using Grammatical Framework in Validation

The GF is a powerful tool to generate and parse domain-spcific, regular sentences and phrases. [x]

In this experiment, we use GF to produce the validation view.

(A sample of the validation view)

Further plans

  • Using known term harvesters

Results

(no exactly comparable baseline method)

Conclusions

Topic revision: r5 - 2011-08-02 - SeppoNyrkko
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback