Transcript Dias nummer 1

Ontology-based search and
knowledge sharing using
domain ontologies
Sine Zambach,
PhD student, Roskilde University
GERPS ‘08
Outline
1. Why Domain Ontologies?
2. Ontology-based search
3. Domain analysis: Relations in
ontologies
4. How does this gain value for the
organisation?
2
Why Domain Ontologies?
Knowledge sharing for common
understanding in e.g. software
development and translations
Background for domain specific
information retrieval
3
Example of a domain ontology
4
Ongoing example
substance
process
isa
isa
Insulin
activates
Glucose
uptake
5
Ontology-based search
Ontology background for
information retrieval:
Broaden search wrt synonyms,
ontological similarity, relations, etc.
Can potentially be used by
organisations to search through all
kinds of texts
6
Ongoing example
substance
process
isa
isa
Insulin =
INS
activate
activate
Glucose uptake =
Glycose transport
New unknown
substance
7
Ontology based search in
biomedical texts
Siabo project
Computer scientists computational
linguists, domain experts,
terminologists
Develops
 Background ontology
 Text preprocessing tools
 Knowledge extraction tools
 Implementation on the texts
8
The SIABO-project
Ontology based search application
Computational Linguists
(CL)
Knowledge Engineers (K)
Computer Scientists (CS)
Terminologists (T)
Domain experts (D)
Knowledge extraction
Search implementation
Text pattern rule
development on NP’s (CL,
KI, D)
Interface (CS)
Search functions (CS, K, D)
Similarity measures (CS)
Text preprossesing
Domain ontology modelling
Grammatical parsing/ POS-tagging/ (CL)
Grabbing/ontological tagging fragments using
ontotypes (K)
Mapping into ontology (CS)
Indexing (CS)
Start from UMLS (T,D)
Modeller in a suitable tool (T,D)
Put into relational database (CS)
9
Relations
Semantic glue between concepts
(the idea behind words)
General and domain specific
relations
Represented by e.g. verbs and can
be identified in various ways
Parallel to concepts that are
represented by terms
10
Relations as semantic ”glue”
Insulin activates glucose uptake
Pancreas activates organ (odd)
Substance activates substance
Substance activates process
11
Domain specific relations
12
OBO-ontologi
Table 3
Some properties of the relations in the OBO Relation
Ontology
RelationTransitive Symmetric Reflexive Antisymmetric
is_a
+
+
+
part_of
+
+
+
located_in
+
+
contained_in
adjacent_to
transformation_of +
derives_ from +
preceded_by +
has_participanthas_agent
Smith et al. Genome Biology 2005 6:R46
13
Domain specific relations
Inhibition and activation
Domain specific Bio-relations
Has interesting properties through a
path of relations of that types.
The relation of ”activation” is
transitive, where ”inhibition” is
more complex and is dependent of
the stimulation-relation
14
Example: positive relation –>
transitivity?
A activates B
B activates C
-> A activates C
A
B
C
A
B
C
A
B
C
15
Example: inhibits and
stimulate -> complex property
A inhibits B
B inhibits C
-> A activates C
A
B
C
A
B
C
A
B
C
16
Verb frequences in the 4
corpora:
17
Background
18
Relations in an enterprise
ontology
Discovering of weird words =
domain specific concepts and
relations
Similarity measure in information
retrieval
Information fishing of new concepts
19
Ongoing example
substance
process
isa
isa
Insulin =
INS
activate
activate
Glucose uptake =
Glycose transport
New unknown
substance
20