Final Activity Report Summary - UMDM (UMDM: Uniform Management of Data and Meta-Data)
We are witnessing a tremendous proliferation of database sources in many aspects of our lives. These sources contain large volumes of information and are heterogeneous in their design and content. To make a meaningful use of this information, modern information systems need to be able to understand process, query, integrate and maintain it. In successfully performing these tasks, metadata plays an important role. Meta-data examples include schema information, data constraints, user comments, ontologies, quality parameters, data annotations, provenance information, etc. Unfortunately, existing database systems do not provide true support for meta-data. This forces data administrators to manually perform the metadata management task which is laborious, time consuming and error-prone. Although there are a few systems developed for that goal, they are usually ad-hoc applications and are limited to only certain kinds of metadata.
In this work we have developed techniques and tools for efficiently and effectively managing data and metadata. This direction had been driven by two main observations. The first is the fact that the volume and the different kinds of metadata have grown extremely large and manual techniques will soon become practically impossible to apply. The second is that the distinction between data and metadata has been blurred to the point where the same piece of information is seen as data by some people and as meta-data by others. We have developed a system for the uniform management of the different kinds of metadata and we have extending existing query languages with primitives to support query and retrieval functionality on it.
The mechanism we provided allows not only the modelling of the data, but also the ability to define associations between data in a declarative way. This way, one can use our framework on database even if he/she has no write access. Furthermore, the mechanism is able to handle future data, i.e. data that is not currently available in the database but which may appear in the future. We have also concentrated on a special kind of meta-data which is called mappings. Mappings are expressions that specify how data instances in different repositories related to each other. Mapping definition is a time consuming and error-prone task, so mapping generation tools have been developed to assist the user in that task. Unfortunately, till today there is no universally accepted method to evaluate and compare these tools. Thus, we have developed the first benchmark for that purpose. The benchmark is freely available on the web at http://www.stbenchmark.org.
In this work we have developed techniques and tools for efficiently and effectively managing data and metadata. This direction had been driven by two main observations. The first is the fact that the volume and the different kinds of metadata have grown extremely large and manual techniques will soon become practically impossible to apply. The second is that the distinction between data and metadata has been blurred to the point where the same piece of information is seen as data by some people and as meta-data by others. We have developed a system for the uniform management of the different kinds of metadata and we have extending existing query languages with primitives to support query and retrieval functionality on it.
The mechanism we provided allows not only the modelling of the data, but also the ability to define associations between data in a declarative way. This way, one can use our framework on database even if he/she has no write access. Furthermore, the mechanism is able to handle future data, i.e. data that is not currently available in the database but which may appear in the future. We have also concentrated on a special kind of meta-data which is called mappings. Mappings are expressions that specify how data instances in different repositories related to each other. Mapping definition is a time consuming and error-prone task, so mapping generation tools have been developed to assist the user in that task. Unfortunately, till today there is no universally accepted method to evaluate and compare these tools. Thus, we have developed the first benchmark for that purpose. The benchmark is freely available on the web at http://www.stbenchmark.org.