Year: 2007
Type: article
Abstract: Annotations and provenance data play a key role in understanding and curating scientific databases.However, current database management systems lack adequate support for managing annotations and provenance data including:(1) handling annotations at multiple granularities, i.e., at the table, tuple, column and cell levels, (2) propagating annotations along with query answers, (3) querying data based on their annotations, and (4) providing declarative ways to add, archive, and restore annotations.In this paper, we propose to treat multi-granular annotations and provenance as first class objects inside the database.We introduce the concept of "Annotated Relations" along with new operators and extended semantics for the standard relational oDerators in support of annotated relations.We present an expressive and declarative extension to SQL to support the processing and querying of annotated tables.We study several schemes for storing and indexing annotations based on annotation granularity and annotation size.Extensions to PostgreSQL are introduced to support annotated relations and implementation challenges are discussed.Performance analysis illustrates the potential of annotated relations as they achieve up to an order-of-magnitude reduction in storage and I/O costs.
Institution: Purdue University West Lafayette
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2140764807