The following principles for corpus design have been adopted with respect to the compilation of BulNC:

1. Task-independent design ensuring as many monolingual and multilingual data as possible, illustrating different media types with their styles, genres, and domains.


2. Extensibility of the corpus through the inclusion of newly emerging categories attested in language production.


3. Flexibility and robustness of the design in order to facilitate reconsideration and restructuring of classificatory information about the texts. Carefully designed mechanisms for reorganising should ensure that already included texts are not misclassified after the changes.


4. Adoption of mechanisms for accommodating texts that belong to multiple categories while any additional information is also properly stored and remains accessible.


5. Easy access to the relevant documents, including simple and efficient extraction of information, as well as grouping and regrouping of texts into subcorpora.


This corpus design is proposed in order to maintain simultaneously monolingual and multilingual parallel corpora and allow them to be compiled, preprocessed, annotated, evaluated and accessed through common or compatible tools, compliant with metadata and annotation description schemes, as well as with common (or convertible) annotation tagsets. This approach ensures standardisation, reusability and automation at all stages of corpora development and usage.


A uniform framework has been developed for structuring BulNC, data storage format and description of the texts. The corpus design requires a clear-cut structure based on an explicit description of sample categories and explicit mapping between parallel samples in different languages. On the other hand, the corpus structure has to be flexible enough to allow for reorganisation around different categories or languages. This is ensured by a detailed and consistent metadata documentation of corpus samples.