Preliminary Recommendations

 

SGML DTD for a formal description

A SGML DTD (Document Type Definition) is a physical model of the grammar that describes the mark-up of data. When translating our conceptual E-R model to a DTD we have adopted a systematic method, designed in particular to encode most of the constraints, cardinality constraints at least. The principles of transposition from the E-R formalism into SGML are as follows:

• E-R entities become SGML elements.
• Attributes of E-R entities become attributes of SGML elements. When the values of an attribute are mutually exclusive and when they constitute a closed vocabulary, then these values are represented in the form of listed SGML attributes.
• Relationships carrying no attributes, pointing at a shared E-R entity, are expressed by reference links between SGML elements (IDREF or IDREFS depending on the cardinality).
• Relationships that carry attributes are expressed in the form of SGML attribute-carrying elements, connected (by reference link) with the SGML elements translating the E-R entities.

The model expressed at conceptual level by means of the Entity-Relationship formalism is now translated to a SGML DTD that describes a possible physical format of a file containing a lexicon conformant with the model, following the transposition principles we adopted. Thus, although we do not propose a complete E-R model of the lexicon, we here propose a subset of a complete DTD that could encode the lexicon and that could include this subpart to deal with syntactic information. On its own, it is not usable, because it has to be inserted in a global representation of the lexicon. As EAGLES is producing recommendations rather than a complete and finished model, our work can be seen as an example of a possible SGML formalisation, not as a strong constraint.

To make it easier to understand a SGML DTD, we note:

• A SGML ELEMENT is here the expression of something like a structured object
• A SGML ELEMENT has attributes, and in particular the special attribute id, which is a unique key that makes it possible to refer to ELEMENTS by pointing to them from other ELEMENTS through their reference, using attributes that are of type IDREF (a kind of pointer).

 
• Partial SGML DTD