Storing and Querying Historical Texts in a Relational Database

This paper describes an approach for storing and querying a large corpus of linguistically annotated historical texts in a relational database management system. Texts in such a corpus have a complex structure consisting of multiple text layers that are richly annotated and aligned to each other. Modeling and managing such corpora poses various challenges not present in simpler text collections. In particular, it is a difficult task to design and efficiently implement a query language for such complex annotation structures that fulfills the requirements of linguists and philologists. In this report, we describe steps towards a solution of this task. We describe a model for storing arbitrarily complex linguistic annotation schemes for text. The text itself may be present in various transliterations, transcriptions, or editions. We identify the main requirements for a query language on linguistic annotations in this scenario. From these requirements, we derive fundamental query operators and sketch their implementation in our model. Furthermore, we discuss initial ideas for improving the efficiency of an implementation based on relational databases and XML techniques.