Web:http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.84.7967&rep=rep1&type=pdf
Abstract.
EU-IST Integrated Project (IP) IST-2003-506826 SEKT
Deliverable D2.1.1 (WP2)
This deliverable presents an SVM-based algorithm for IE and experiments on several benchmark datasets. The results showed that our system is comparable to other state-of-the-art systems on both traditional IE and adaptive IE tasks. We investigated two feature weighting schemes, the impact of different NLP features on the performance of the learning algorithm, and the importance of the SVM parameters. Results are reported both on traditional IE tasks such as named entity recognition and slot filling, and on adaptive IE and hierarchical named entity learning. Directions for future work are discussed in the conclusion.
Keyword list: Ontology-based Information Extraction, Machine Learning, Adaptive IE