Abstract

In this paper, we want to retain our experiences by participating in TREC 2007 Blog Track 'Feed Distillation'. To perform the run we combine various classifiers analyzing title-, content- and splog-specific features to predict the relevance of a feed related to a topic, based on the idea of AdaBoost. The implemented classifiers are based on keywords retrieved from different thesauri such as Wordnet and Wortschatz, as well as websites providing hierarchical organized 'ontology' such as the 'Open Directory Project' and Yahoo Directory. To structure the keywords, we use Topic Maps.

@inproceedings{Lee2008,
   author = {Wai-Lung Lee and Andreas Lommatzsch and Christian Scheel},
   title = {Feed Distillation Using AdaBoost and Topic Maps},
   booktitle = {The Sixteenth Text REtrieval Conference (TREC 2007) Proceedings},
   year = {2008},
   doi = {http://trec.nist.gov/pubs/trec16/papers/techu-berlin.blog.final.pdf},
   editor = {E. M. Voorhees and Lori P. Buckland}
}