Abstract: A computer apparatus which uses a database (22) to offer persistent storage of metadata (36) describing the content of media files (32). Metadata is used to create a personalised media article (48) from those media files. That metadata also indicates relationships between those media files. In order to accelerate the creation of the personalised media article (48), media element metadata items are stored in a cache. The usefulness of this cache is improved further by reading related media elements data from retrieved metadata items (36), and then pre-fetching those items and placing them in the cache. Because the relatedness of the data items means that the related data item is more likely to be required in the near future, the caching method is more useful than known caching methods. Furthermore, the improved usefulness of the cache is achieved without placing constraints on where the data items are placed in the persistent storage.