Title Extraction of Hyperonymy of Adjectives from Large Corpora by Using the Neural Network Model 
Author(s) Kyoko Kanzaki (1), Qing Ma (2), Eiko Yamamoto (1), Masaki Murata (1), Hitoshi Isahara (1)

(1) National Institute of Information and Communications Technology; (2) Ryukoku University

Session P5-W
Abstract In this research, we extract hierarchical abstract concepts of adjectives automatically from large corpora by using the Neural Network Model. We show the hierarchies on the Semantic Map and compare the hierarchies in the Semantic Map and a manually prepared thesaurus. We recognized five types of distributions on the map. By comparing the Semantic Map and a manual thesaurus, we found that the word that the abstract noun belongs to, whether a person, thing or event, is introduced as the standard of classification in the manual thesaurus. On the other hand, in the Semantic Map, we found that abstract nouns belonging to people or events are distributed together. We also found that the hierarchies of sokumen (side), imi (meaning), and kanten (viewpoint) are necessary for a category of adjectives.
Keyword(s) Hierarchical relations, Neural Network Model, Abstract Nouns, Adjectives
Language(s) Japanese 
Full Paper 616.pdf