A hybrid information retrieval model using metadata and text

Abstract

Metadata is data about data. It means that metadata describes the property of data generally. In other words, metadata provide information related to the raw data such as video, audio, text and so on. By etadata, we can find information not only easily and fast but also based on semantics and non-topical attributes. These are the reasons why various metadata are created and studied. To retrieve information based on metadata fields effectively, however, we should know the characteristics and structure of metadata. Commonly, most users are not familiar with constructing a query for metadata retrieval. Therefore, it is necessary to translate a user information need to a query for metadata automatically. The amount of information in metadata is usually less than like in the full data like text. If a user wants information not described in a metadata field, the user can never find relevant information. For instance, when we cannot find relevant information only by searching the metadata, there is a chance that it can be found in the text. In this paper, we suggest a hybrid IR model using metadata and text which can provide users with additional relevant documents by searching the metadata fields and text fields. User queries are translated into structured or unstructured queries automatically to search the metadata fields and text fields simultaneously, even though the user does not understand what metadata is.