Automatic Biological Term Annotation

Exciting research in biology has resulted in a large
amount of biological publications.
Knowledge discovery in biology becomes an interesting
task which can be established
by recognizing terms in text to extract useful
information such as interaction relationships.
We propose the Automatic Biological Term Annotation
(ABTA) system which uses classification methods to
annotate terms in text. A novel method is presented
to express lexical features in pattern notations.
Prefix and suffix characters are used instead of
lists of potential terms or external resources. We
demonstrate that part-of-speech tag information is
the most effective attribute. Creating classification
exemplars is conducted from text by using word n-gram
model. We illustrate improvements on our system''s
performance which depends on the feature attributes
we define. Biological concept markers are also
assigned to each located term indicating its meaning.
Our results are comparable to the performance of
other existing systems while our system retains
simplicity and generalizability.

Autorentext

Sittichai Jiampojamarn is a PhD student at the Department of Computing Science, University of Alberta, Canada. He received his Master of Computer Science degree from Dalhousie University, Canada in 2005. His research interests are in Natural Language Processing (NLP), Machine Learning (ML), Information Extraction and Bioinformatic research areas.

Klappentext

Exciting research in biology has resulted in a largeamount of biological publications. Knowledge discovery in biology becomes an interestingtask which can be establishedby recognizing terms in text to extract usefulinformation such as interaction relationships.We propose the Automatic Biological Term Annotation(ABTA) system which uses classification methods toannotate terms in text. A novel method is presentedto express lexical features in pattern notations.Prefix and suffix characters are used instead oflists of potential terms or external resources. Wedemonstrate that part-of-speech tag information isthe most effective attribute. Creating classificationexemplars is conducted from text by using word n-grammodel. We illustrate improvements on our system'sperformance which depends on the feature attributeswe define. Biological concept markers are alsoassigned to each located term indicating its meaning.Our results are comparable to the performance ofother existing systems while our system retainssimplicity and generalizability.