[01:29:18] Congrats on a finished phase epsilon! [01:29:19] Of interest to those in this chat might be some proposed qualifiers for marking up syntactic relationships: https://www.wikidata.org/wiki/Wikidata:Property_proposal/dependency_grammar_relations [02:27:17] Wow, interesting [05:51:50] To my knowledge none of the five focus languages has extant UD annotated corpora, so this is certainly an interesting area to explore for them [15:49:52] Reading up on UD (a fairly new thing in last 6 years evidentally) in https://en.wikipedia.org/wiki/Universal_Dependencies#Function_words and looking at that first structural analysis diagram that UD would represent, it seems that it would be very close to what I would expect a Constructor to pull together. "say likes swim" is very short, but gives the general or "Abstract" meaning. UD then pushes down the importance of sy [15:49:52] This is exactly how I think when I'm envisioning some functions and constructors within Abstract Wikipedia. To quote old police movies - "Give me just the facts ma'am" ๐Ÿ˜ The WHAT parts only. "say likes swim". [15:49:54] Later on, through better dependency analysis, it's easier to take the hanging parts and map the dependencies to fully form longer better sentences that layer in the rest of the context. Who, WHAT, When, Where, How [15:50:24] Reading up on UD (a fairly new thing in last 6 years evidentally) in https://en.wikipedia.org/wiki/Universal_Dependencies#Function_words and looking at that first structural analysis diagram under Function words that UD would represent, it seems that it would be very close to what I would expect a Constructor to pull together. "say likes swim" is very short, but gives the general or "Abstract" meaning. UD then pushes down [15:50:25] This is exactly how I think when I'm envisioning some functions and constructors within Abstract Wikipedia. To quote old police movies - "Give me just the facts ma'am" ๐Ÿ˜ The WHAT parts only. "say likes swim". [15:50:27] Later on, through better dependency analysis, it's easier to take the hanging parts and map the dependencies to fully form longer better sentences that layer in the rest of the context. Who, WHAT, When, Where, How [15:52:09] Reading up on UD (a fairly new thing in last 6 years evidentally) in https://en.wikipedia.org/wiki/Universal_Dependencies#Function_words and looking at that first structural analysis diagram under Function words that UD would represent, it seems that it would be very close to what I would expect a Constructor to pull together. "say likes swim" is very short, but gives the general or "Abstract" meaning. UD then pushes down [15:52:10] This is exactly how I think when I'm envisioning some functions and constructors within Abstract Wikipedia. To quote old police movies - "Give me just the facts ma'am" ๐Ÿ˜ The WHAT parts only. "say likes swim". [15:52:12] Later on, through better dependency analysis, it's easier to take the hanging parts and map the dependencies to fully form longer better sentences that layer in the rest of the context. Who, What, When, Where, How [15:52:29] Reading up on UD (a fairly new thing in last 6 years evidentally?) in https://en.wikipedia.org/wiki/Universal_Dependencies#Function_words and looking at that first structural analysis diagram under Function words that UD would represent, it seems that it would be very close to what I would expect a Constructor to pull together. "say likes swim" is very short, but gives the general or "Abstract" meaning. UD then pushes dow [15:52:30] This is exactly how I think when I'm envisioning some functions and constructors within Abstract Wikipedia. To quote old police movies - "Give me just the facts ma'am" ๐Ÿ˜ The WHAT parts only. "say likes swim". [15:52:31] Later on, through better dependency analysis, it's easier to take the hanging parts and map the dependencies to fully form longer better sentences that layer in the rest of the context. Who, What, When, Where, How [15:53:35] That's similar to what I was thinking regarding its usability in a renderer, although this was admittedly secondary to "lots of languages have corpora in it, so we have diverse examples to look at" [15:54:31] It mentions the UD annotation scheme as controversial, but I just don't see that. To me it gives better structure of dependencies than traditional analysis which concerns less with semantics and more with syntactical considerations. [15:57:12] Yea, well, I'm not that concerned about corpora in general, there's thousands out there. More concerning is us continuing to find what we actually need for Abstract Wikipedia's mission which is finding "high quality, quickly useful annotated corpora". [15:58:16] "say likes swim" fits perfectly in something like a "Base Simple English" ... informally known as "Baby English" ๐Ÿ˜Š [15:59:09] Yea, well, I'm not that concerned about corpora in general, there's thousands out there. More concerning, as you stated above is us continuing to find what we actually need for Abstract Wikipedia's mission which is finding "high quality, quickly useful annotated corpora". [15:59:18] thousands in English, which I will reassert is not the center of the lexicographical universe (re @thadguidry: Yea, well, I'm not that concerned about corpora in general, there's thousands out there. More concerning, as you stated above is us continuing to find what we actually need for Abstract Wikipedia's mission which is finding "high quality, quickly useful annotated corpora".) [15:59:23] Yea, well, I'm not that concerned about corpora in general, there's thousands out there. More concerning, as you stated above, is us continuing to find what we actually need for Abstract Wikipedia's mission which is finding "high quality, quickly useful annotated corpora". [15:59:41] A GREEEEED [15:59:51] Now that would be an interesting variety of English to provide renderers for (re @thadguidry: "say likes swim" fits perfectly in something like a "Base Simple English" ... informally known as "Baby English" ๐Ÿ˜Š) [16:03:23] hmm, does GF let me create "Baby English" easily? does it give me quick UD analysis to create "Baby English? [16:05:34] Hmm, yeah, looks like it, if you just have it focus output custom trees with only Kind, and ... ? hmm... any of the V's ? https://www.grammaticalframework.org/doc/tutorial/gf-tutorial.html#toc21 [16:05:36] @mahir256 do you happen to know already? [16:05:57] Nope [16:06:08] no idea [16:06:21] I'll ask their community then [16:16:38] Found it I think... only the "Open classes" in the Lexical categories. Also known as the "content words", and not the "function words". https://www.grammaticalframework.org/doc/tutorial/gf-tutorial.html#toc73 [16:18:18] "this pizza is good" -> BABY_ENGLISH_RENDERER() -> "pizza good" [16:20:42] -> "kรคse kรถstlich" [16:29:05] Awesome! They already do the UD analysis work for me! And it works with the "vd" command in GF [16:29:06] https://github.com/GrammaticalFramework/gf-ud [16:31:42] > At the time of writing, ud2gf can get very slow and resource-hungry. Its main usage is for the conversion of small UD trees (less than 10 words) into GF linearization rules. [16:33:20] I'd like renderers to form UDish syntax trees first and then convert those into strings, rather than convert from something else into such a representation later [16:33:21] but seems from what folks, the other way around... gf2ud is decent. [16:33:39] but seems from what folks say, the other way around... gf2ud is decent. [16:34:01] yeah, I hear you (re @mahir256: I'd like renderers to form UDish syntax trees first and then convert those into strings, rather than convert from something else into such a representation later) [16:38:07] > At the time of writing, ud2gf can get very slow and resource-hungry. Its main usage is for the conversion of small UD trees (less than 10 words) into GF linearization rules. [18:11:48] What's UD? [18:12:34] read the property proposal :-) there's links to it there (re @jhsoby: What's UD?) [18:12:55] ("Universal Dependencies", though that name by itself doesn't say much)