[02:34:52] I love this vision, @DennisPriskorn [02:35:15] Some of the use cases you describe require parsing, and that's very hard [02:36:26] But having functions integrated directly in LibreOffice, that would be great [02:36:57] And providing government texts like laws or tax codes in abstract notation, yes that would be pretty awesome [02:37:39] I'm particularly thinking legal texts, medicine texts, and scientific texts [02:38:31] I like this kind of dreaming. Whether we can get this far, and whether we can reach sufficient coverage, that'll be the interesting question for the next few years [02:39:08] Hence why adding syntactic qualifiers to combines on multi-part lexemes is so important (re @vrandecic: Some of the use cases you describe require parsing, and that's very hard) [02:39:37] That's also important for generation! [02:41:10] all text entered into computers being matched to lexemes seems completely unrealistic to me though. that would require language to be a finite fixed thing, which it isn't [02:41:51] Yes. That's why I'm focusing on the very limited generation of encyclopedic articles. [02:42:05] And not just the vocabulary, the grammar! [02:42:58] The way most annotation systems work is that you always have a "unknown token" annotation. Yes, it's cheating [02:48:50] I found it amusing that the Slate article on Abstract Wikipedia had the seeming neologism "reinforced". I loved that word, as I found the meaning immediately clear in the context even though it was new. Something you won't be able to do with Abstract Wikipedia. I see now that the text was changed to say "ring-fenced" [02:51:42] huh? but "reinforced" has existed for centuries? [02:52:21] Reinfenced [02:52:28] Autocorrect! [02:59:18] indeed it has (since 1485 as a part participle, since 1612 as an adjective, according to the OED) (re @Nikki: huh? but "reinforced" has existed for centuries?) [06:43:51] Exactly and for the cases that are ambiguous, I think a dialog similar to spell check is needed. We could call it semantic ambiguous check 😃 (re @vrandecic: Some of the use cases you describe require parsing, and that's very hard) [06:45:08] Exactly and for the cases that are ambiguous, I think a dialog similar to spell check is needed. We could call it semantic ambiguous check 😃 [06:45:09] The idea is than an IA assists in matching most of the words to lexemes and the rest have to be done manually. (re @vrandecic: Some of the use cases you describe require parsing, and that's very hard) [06:45:19] Exactly and for the cases that are ambiguous, I think a dialog similar to spell check is needed. We could call it semantic ambiguous check 😃 [06:45:19] The idea is than an AI assists in matching most of the words to lexemes and the rest have to be done manually. (re @vrandecic: Some of the use cases you describe require parsing, and that's very hard) [06:50:16] Yeah, it's a challenge, people are creating hundreds of new words and combinations every day, only a small subset ever reaches publication I guess. So for creative stuff it's probably now worth the effort, but in the domains Denny mentioned it might prove very valuable. (re @Nikki: all text entered into computers being matched to lexemes seems completely unrealistic to me though. that would require language to be a fini [06:50:31] Yeah, it's a challenge, people are creating hundreds of new words and combinations every day, only a small subset ever reaches publication I guess. So for creative stuff it's probably not worth the effort, but in the domains Denny mentioned it might prove very valuable. (re @Nikki: all text entered into computers being matched to lexemes seems completely unrealistic to me though. that would require language to be a fini [17:07:40] I would love for that to happen!