The reason for complete rewrite was simple:
Это отвлекло нас от нашего спора.
That distracted us from our argument.
Every word makes a contribution to the parsing state by saying that it participates in a number of constructions, and provides some attributes for each of them. So a contribution consists of pairs of constructions and attribute sets. I'll call such pairs mites.
That (это) is capable to add noun attributes to nom(inative) and acc(usative) constructions, and these mites contradict each other. Distracted (отвлекло) is a transitive verb, therefore it defines head for nom and acc constructions. But these two mites don't contradict each other, they're both very welcome.
Earlier, I unified mites as soon as they came, and I only kept the results of the unification, not the mites themselves. After the word that, there were nom and acc constructions with noun defined, and nom was (randomly) chosen as active. Active constructions had higher priority, and therefore the verb's nom mite was integrated first. Two nom mites were unified, and the nom construction with both head and noun defined linked the values of these attributes semantically. That's all great. But what to do with acc?
You can't just unify it as well, because there's already a nom construction and it contradicts acc in its first mite. So you have to drop either of the mites completely and leave the other. In this sentence, it made sense to drop the acc.noun mite — the accusative alternative of that. The acc.head mite would survive and unify nicely with the real object that comes next — us. But that's a hack and that didn't work when the first это was in fact accusative. A more right way would be to preserve both acc mites but only mark the second one as active. That clearly required per-mite active status, not per-construction.
So now I don't remove any mites automatically when processing word contributions, they all survive. The matter is, are they active or not. Active mites are guaranteed to be non-contradicting. Only active mites about the same construction are unified and passed to this construction, it then may make some semantic changes based on that information.
When another word comes, the parser solves a constraint satisfaction task on graphs. Some of the new mites are chosen to be active, the old mites may change their active status to accommodate that. The reanalysis tries to
- minimize the number of active constructions
- to maximize the number of mites in those constructions (more mites mean more information)
- prefer more recent mites
- to prefer semantically more plausible variants
Surprisingly, it works, though twice as slow as before. And the algorithm now seems so clean and logical that I can't find anything to remove! But I definitely should optimize, as I just get bored waiting for long 10 seconds until 171 tests pass.