Automatic Taxonomy Construction for Eye Colors Data without Using Context Information
Jing Qiu, Yaqi Si, Zhihong Tian · 2018
Taxonomy Learning is one of the challenge subtasks of Ontology Learning. A taxonomy describes hyponymy between concepts, and it plays an important role in information systems and categorization tasks. Works in this field prefer to utilize context of sentences to obtain a precise analysis of the corpus, since context can provide more semantic information for building a precise interpretation of the semantic classes. In this paper, we propose a method, called SenseRefined, for automatic taxonomy construction without using context information. It's a part work of one of our project. The proposed method is specially applied for eye colors data, and all the domain phrases are extracted out before building taxonomy. Word senses are extracted out as a way to provide semantic information, because we find the senses of words are more like the definitions of the concepts, which often contain hypernym concepts in. Experiments were carried out using SenseRefined method and Substring Matching method, where the latter one is baseline. Experimental results show the effectiveness of our approach.