LivingKnowledge goal is to bring a new quality into search and knowledge management technology for more concise, complete and contextualised search results.
The paper “Towards Using Reranking in Hierarchical Classication” co-written by Q. Ju, R. Johansson, and A. Moschitti has been presented In the proceeding of Second Pascal Challenge on Large Scale Hierarchical Text Classification, in Athens, Greece, 2011.
We consider the use of reranking as a way to relax typical independence assumptions often made in hierarchical multilabel classication. Our reranker is based on (i) an algorithm that generates promising k-best classication hypotheses from the output of local binary classiers that classify nodes of a target tree-shaped hierarchy; and (ii) a tree kernel-based reranker applied to the classication tree associated with the hypotheses above. We carried out a number of experiments with this model on the Reuters corpus: we rstly show the potential of our algorithm by computing the oracle classication accuracy. This demonstrates that there is a signicant room for potential improvement of the hierarchical classier. Then, we measured the accuracy achieved by the reranker, which shows a signicant performance improvement over the baseline.