Hyponym Extraction from the Web based on Property Inheritance of Text and Image Features

Concept hierarchy knowledge, such as hyponymy and meronymy, is very important for various Natural Language Processing systems. While WordNet and Wikipedia are being manually constructed and maintained as lexical ontologies, many researchers have tackled how to extract concept hierarchies from very large corpora of text documents such as the Web not manually but automatically. However, their methods are mostly based on lexico-syntactic patterns as not necessary but sufficient conditions of hyponymy and meronymy, so they can achieve high precision but low recall when using stricter patterns or they can achieve high recall but low precision when using looser patterns. Therefore, we need necessary conditions of hyponymy and meronymy to achieve high recall and not low precision. The previous papers have assumed “Property Inheritance” from a target concept to its hyponyms and/or “Property Aggregation” from its hyponyms to the target concept to be necessary and sufficient conditions of hyponymy, and proposed several methods to extract hyponymy relations from the Web based on property inheritance and/or property aggregation of text features such as meronyms and behavior. This paper proposes a method to acquire hyponymy relations from the Web based on property inheritance of not only text features but also image features for each conceptual word. Keywords-hyponymy; meronymy; concept hierarchy; Web mining; image analysis; property inheritance; typical image;