Abstract
In recent years, the number of articles that describe protein structures and functions is rapidly increasing. Retrieval of desired articles from a large collection of articles is highly needed. This paper presents an article-based retrieval method for articles on protein structure analysis. In this method, an article is considered as a set of structural and functional concepts of protein using Gene Ontology and other databases, and the similarity between articles can be evaluated by using a concept graph generated by integrating Gene Ontology and InterPro database. An article includes various types of information, which makes it difficult to decide the viewpoint from which the users intend to retrieve articles in the article-based retrieval. To solve this problem, our proposed method use more than one article as a single query to clarify users' intention in the retrieval request,which improves the retrieval accuracy. The effectiveness of our proposed method was confirmed by evaluating its accuracy through retrieval experiments, especially for retrieving more recent articles.
