Amir Plivatsky

In order to implement the viterbi algo for parsing, I first need to learn it more deeply... For actually implementing it, I will need more time since I would like...

> 1) If you want to add an alternative constraint to the existing code, that's fine by me, as long as it makes things mostly faster. This is something I...

I still need to understand more the concept of "dynamic left-wall"...

I just tried pruning alternatives (because it is easy to try), and didn't observe a speedup even for Russian (`ru/corpus-basic-batch` concatenated 50 times). This is expected because `form_match_list()` must still...

> [...] and became confusing. Their English is indeed somewhat sloppy, and unintended words are used in a significant frequently... However, I only used the results as summed up in...

> There's also a different issue: the way that E_lists are used is kind-of crazy: the way that they are and'ed, or'ed, and the way that null connectors (optionals) get...

> In theory, someday, some C compiler might be smart enough to do this automatically; compiler designers often try to think of ways to avoid having to create new stack...

> After the fix, maybe other optimizations would become clear? Assuming the fix is not a performance de-accelerator? I wanted to check that. However, the "**lr**" branch is not in...

> , nor how the existing E_list design was arrived at. While constructing the expression iterating functions (a WIP that I also don't know if it will turn out to...

I wrote also a stack-based non-recursive purge_Exp(). It seems to make a perfect job (checked with `link-parser v=9 -debug=exprune.c` with an input batch file, compiled with debug and ASAN/LSAN/UBSAN, and...