Maarten van Gompel
Maarten van Gompel
Ko heeft de meeste vragen al beantwoord zie ik. FoLiA is inderdaad erg verbose en kan snel een bottleneck zijn als je echt heel veel data hebt. * 1: Dat...
> Moet ik die 'Compressed binary representation' als zoiets als huffmann coding > zien? Of meer een platte index zoals hierboven beschreven, ik zal zo ook even > naar de...
Yes, agreed
Another good (secondary) source for location data is OpenStreetMap, I experimented with that yesterday. It's fairly easy to extract all streets and cities/towns.
This is also relevant for @Irishx (frog evaluation) and @HenkvdHeuvel (112 project), and perhaps @antalvdb: Okay, things are a bit more complex. We have some ordering problems. The current situation:...
Dat lijkt me een goed idee ja, je kan de gazetteers in ieder geval uitschakelen door zelf in de frog configuratie file, en ``ners.known`` te editen.
I like this idea (and am open to pull requests). I'm not sure about the syntax yet though. It's a bit too convoluted to integrate it into --showcolumns, we'd better...
Thanks, but I think it's up to the various Linux distributions/packagers to provide binary packages and supply them through their ecosystem.
Sorry, I can't really answer this question as I'm not familiar enough with the actual models, but perhaps @ltenbosch or @HenkvdHeuvel can comment on this.
I'm not entirely sure actually how extensible these models are once trained, they're don't work like wav2vec. Perhaps @ltenbosch can answer this question better?