A hashing algorithm based on tagging a piece of knowledge with some descriptive “head” (key) which is then organised alphabetically, augmented by including the first vowel after the initial letter. The unaugmented alphabetical approach results in an unbalanced table (i.e. wasted paper), the use of the vowel helps distribute the heads more evenly. Also by a quirk of language also helps arrange concepts with similar etymological roots side by side.
What I don't exactly get is how the pages are assigned. It's not really a hashing algorithm in that sense, it's more like a dictionary data structure where the user has to choose the storage locations. It's only possible to work from one end or the other, or more or less randomly. Either way fragmentation will creep in. This looks like it happens in the image of Locke's index.