5 Conclusions

At this stage of the DLMF project, we are still very much focused on obtaining the content. Much of the effort up to this point has gone into TEX programming to provide support for a subtly modified level of L^ATEX markup. The goal is to allow authors to minimize presentation markup and ambiguity, while maximizing content markup. We try to enable the authors to concentrate on the mathematics and develop the content as painlessly as possible. The ulterior motivation for this is that the material can then easily be presented in different media: in print and on the web.

We are moving into the next phase, focusing on the transformation to XML and Presentation MathML. This latter, using systems such as tex4ht[7, Ch. 4] seems within reach, although some refinement is still needed.

It is clear that as we move into further stages, parsing into fully Content MathML or OpenMath, additional refinement and markup of the mathematics will be necessary, even adoption of more verbose, unambiguous markup. But, with the approach outlined we should already be relatively close to that state. As long as we understand the author's intent, we will be able to augment the sources with additional markup and declarations to complete the parsing. But while the tools to carry out that parsing are not yet available, applications to use such content markup are also not yet widespread, so the pressure is not yet intense.

Between the metadata markup and modestly semantic mathematical markup, we already have sufficient data for the augmented text-based search techniques described here. We believe that with this approach powerful search capabilities will be available to DLMF users. Our approaches are rather generic, and should be immediately applicable to other mathematics databases. If these methods prove insufficient, and once we have made further progress in capturing the complete structure of the mathematical data, we will look into true mathematical search techniques.