Ontologies have always interested me. As a natural language processing (NLP) researcher, I have worried about linguistic resources for NLP tasks. Therefore, it was a pleasure to read this paper on a system the authors developed as an enhancement to Wikipedia. They describe the language acquisition bottleneck and provide a solution to the problem: a taxonomy derived from Wikipedia that, besides making Wikipedia more powerful and complete, allows for more accurate information retrieval. They also offer a large and comprehensive bibliography, containing references concerned with taxonomy and NLP applications that need such lexical resources.

The authors present research on taxonomy induction based on two lexical resources, Cyc and WordNet. At times, the text can be a bit confusing, due to the large amount of data presented. In addition, the runtime required for taxonomy generation is too long, according to the text. Furthermore, the authors fail to explain how they calculate accuracy. It also would have been very interesting to know the cost and time savings this method affords, compared with other approaches.

The employed methods are clear and well defined, and the evaluation is well presented. Besides its precision, Ponzetto and Strube also prove that their system is innovative. They mention the Kylin ontology generator (KOG) [1], which is similar to their system, and highlight some of their similarities and differences.

Overall, the paper is very good. I recommend it to anyone seeking a comprehensive presentation of many aspects of taxonomy induction based on collaboratively built sources.