Semantic Annotation on the Linked Data Cloud - Experiments
and Evaluation
PRESENTER:
Michel Gagnon
Ecole Polytechnique de Montreal
ABSTRACT:
Semantic annotation, the process of identifying key-phrases in texts and
linking them to some concepts from a knowledge base, is an important basis
for developing semantic information retrieval technologies. Despite the
recent emergence of semantic annotation systems, very few comparative
studies have been published on their performance.
In this talk, I will present two works that result from a collaboration
between Prof. Amal Zouaq, from University of Ottawa, and the WeST lab, at
Polytechnique Montrl. I will first present a semantic annotator that is
based on the principle of collective disambiguation. I will then present
an evaluation of the performance of existing systems over three tasks:
full semantic annotation, named entity recognition, and keyword
detection. The evaluation is achieved in two steps: we first compute
precision and recall on several corpora, and we then build a statistical
model, using logistic regression, to a) identify significant performance
differences and b) provide some estimate of the expected performance of
semantic annotators on similar but unknown corpora.
BIO: Michel Gagnon is a professor at the Computer Engineering
Department of
Polytechnique Montreal since 2002. Previously, he worked as a team leader
at Machina Sapiens inc., a company which at that time was a leader in the
development of grammar checkers, and as a professor at the Univerdade
Federal do Parana, in Brazil. He received his Ph.D. degree in computer
science in 1993 from the Universitde Montreal. Since then, he has been
working on natural language processing, with a special attention to
semantics. Since 2002, his research activities also include the semantic
web, e-learning and security, with a particular interest in the knowledge
extraction from texts.