The principal requirement of a field that will be faceted on is that it must be indexed; it does not have to be stored. And for text fields, tokenization is usually undesirable. For example, if the value Non-Album Track
was tokenized, faceting on a field with that value would show tallies for Non-Album
and Track
separately. On the other hand, tag-clouds, some approaches to hierarchical faceting, and term-suggest are faceting use cases that handle tokenization just fine. Keep in mind that with faceting, the facet values returned in search results are the actual indexed terms, and not the stored value, which isn't used.