Comments from reviewers for GWC 2010

Review 1

This paper provides considerations about the relevance for NLP tasks of traditional manually crafted resources like WordNet and the new sets of structured information derived from Wikipedia. In particular the attention is focused on hypernym hierarchies in WordNet and Wikipedia. From the Japanese Wikipedia a collection of hypernym hierarchies including, as presumable, many named entities has been derived and expanded by exploiting NP modifiers. These hierarchies have been mapped to the Japanese WordNet. Conclusions and remarks about their mappings and their complementarities are provided.

The paper is well-written and structured: it addresses, focusing on the hypernyms hierarchies, important aspects of the relevant issue concerning possible interactions or differences between WordNets and structured data derived from the collaboratively edited Wikipedia, also providing interesting conclusions. It should be better to provide more bibliography considering examples of proposals supporting Wikipedia mined structured-information in substitution of WordNet data (radical enthusiast of Wikipedia) and examples of other attempts to compare these two kinds of resources.

Minor remarks:
- section numbers referred in Paragraph ‘2.1 Organization’ are not correct - footnote 1 at page 2 is not indictated.

Review 2

This paper describes a semi automatic way to extract hypernym hierarchies from Japanese Wikipedia categories. The roots of these hierarchies are then linked to Japanese WordNet leaves. The statistical results of this connection are showed together some final considerations.

This well written paper, in a motivating story, faces a very popular question. Is Wikipedia actually superior over WordNet? Even though the answer could look affirmative the authors demonstrates that WordNet and Wikipedia can become complementary resources. In particular the categories of Wikipedia don't represent a solid upper ontology for the "named entities" constituted by the pages of Wikipedia. We would like to have more information about the linkage process between the hypernym hierarchies extracted from Wikipedia and WordNet leaves. In the article the problem of polysemous words is just mentioned.

Minor remarks
in section 3.3 “quite a log/t of imagery entities”

Review 3

This paper treats very interesting topics, i.e. comparison of hand-crafted thesaurus (WN) and automatically extracted thesaurus (from Wikipedia). Comparison in this paper is objective and understandable. We would like to know how to combine these two kinds of resources to get best result (for example, in some NLP applications).