Mif2Go provides built-in content models for basic DITA topic types. If you are using structured FrameMaker, you can abstract content-model information from your DITA DTD. To add or replace a content model, use free command-line utility dtd2ini to generate a content model from a DTD, and produce a content-model configuration file for your DITA project. See §30.2.2 .
For either structured or unstructured FrameMaker, you must map FrameMaker formats to the content model in terms of element names, element parentage, required level (if any), inline/block nature, and any requirement to be first under any parent. This information goes into configuration file mif2htm.ini, and possibly into chapter-specific configuration files. You might have to use markers in your FrameMaker document to provide information such as topic IDs, element names, and attributes, in cases where these items cannot be derived from the document.
In an unstructured FrameMaker document, presentation might be the same for several different usages. Mif2Go cannot necessarily determine whether (for example) text tagged <Italic> is a computer term, a foreign language term, or a long quote, all of which have different representations in DITA, based on the context. The onus is on the author to disambiguate these usages, if necessary by inserting DITA markers in individual instances of particular formats. Mif2Go does handle a few presentational features automatically; for example, by default forced returns (FrameMaker Shift+Enter) are converted to spaces.