I've created a new MosesPhonemicCollation jar for sorting based on the phonemic representations. I've also forked the dictionary build process based on a parameter called "dictionaryType", which can be "learner" or "linguist". The former produces a dictionary based on the orthography, sorted with the MosesOrthographyCollation, and the latter produces one based on the phonemic transcriptions, with the new collation. The "alphabet" guides that run across the bottoms of pages are also appropriately different. I've abstracted the front matter into a separate file, and I'm auto-including the personography, although I'm not processing it yet.