The FST mechanism for using an acyclic graph to store information is very reminiscent of GADDAG[1] structure for Scrabble dictionaries.
In my search for interesting problems in the compression space, I ran into quite a lot of genomics research which suggests combining a compression algorithm with a lookup method, because identifying common items between samples reduces both storage and speeds up lookup.
> An FST based data structure as presented in this article does not have this capability. Namely, once it is produced, it cannot be changed.
That is somewhat of a weakness in general for succinct data structures of this sort, which makes it harder to remove data (much easier to maintain a "deleted set" to go with it, but that has problems with a delete + re-add).
In my search for interesting problems in the compression space, I ran into quite a lot of genomics research which suggests combining a compression algorithm with a lookup method, because identifying common items between samples reduces both storage and speeds up lookup.
> An FST based data structure as presented in this article does not have this capability. Namely, once it is produced, it cannot be changed.
That is somewhat of a weakness in general for succinct data structures of this sort, which makes it harder to remove data (much easier to maintain a "deleted set" to go with it, but that has problems with a delete + re-add).
[1] - https://en.wikipedia.org/wiki/GADDAG