[03:00:26] are the properties for the objects in wikidata explicitly defined, or are they generated algorithmially? [03:25:33] theology, you mean property IDs, like P31 ? [03:25:52] yurik, nah, like the color of "apple" [03:26:09] ?? [03:26:57] yurik, https://www.wikidata.org/wiki/Q89 [03:27:04] yurik, scroll to "color" [03:27:12] yurik, "color" is not defined for all objects [03:27:22] yurik, was color defined explicitly or algortihmically? [03:27:29] theology, right, color is actually P462 [03:27:35] https://www.wikidata.org/wiki/Property:P462 [03:27:37] yurik, OKAY [03:27:47] users define these properties [03:27:52] thank you [03:27:54] after some discussion [03:28:06] ok, got it, i appreciate your sharing of information with me [03:28:14] and once created, the property gets assigned the next available number [03:28:21] i wonder if these properties will ever be able to be algorithmically defined [03:28:32] yurik, ok [03:28:36] "color" is just a label in english on that object [03:28:57] i am not exactly sure what you mean by algorithmically defined - all of wikidata is (supposedly) human generated [03:29:03] mostly with scripts [03:29:16] so the anthology is fully human controlled [03:29:17] yurik, i mean, with machine learning [03:29:44] yurik, a new object "foo" is created. a machine learning algorithm searches the internet and assigns properties to "foo" [03:30:04] yurik, say, even "foo" is created algorithmically, from a machine learning algorithm searching the internet [03:31:06] well, while it could happen ... eventually... how would that differ from google search? Wikidata is a human-curated data store. Algorithms could potentially generate unlimited amount of data - but then it wouldn't be really possible to curate it - the quality would go down [03:31:46] you basically have a conflict of human generated and maintained, vs machine generated on the fly from a web of unorganized resources [03:31:50] yurik, google searches are not a graph [03:32:03] yurik, its a list of results with description explicitly defined [03:32:08] yurik, dude [03:32:23] sure, but i meant it in a more generic way - algo could produce anything, including a graph [03:32:34] dude. [03:33:13] yurik, a program starts. it searches the internet. assume this program can parse information such that it can create objects and assign properties to these objects based on the information it gathers, and can relate objects to one another, much like the way wikidata does it with subclasses and instances and what not. [03:33:31] yurik, this program algorthimically creates a knowledgebase [03:33:33] not a hard concept [03:33:45] completely different from a "google search" [03:35:11] how is that different? Google creates a knowledge graph, tracing site concept correlations, matches multi-meaning subtrees, and later produces a list from that complex algo. [03:35:38] the point is - its an algo-generated result set. Why store it? [03:36:16] i did not know google creates a knowledge graph on search. [03:36:47] yurik, why not store it? [03:36:56] yurik, imagine if a knowledge base could ask questions [03:37:11] yurik, its facts and questions machine parsable, machine understandable [03:37:36] yurik, it would fill its knowledgebase where there are holes, and discover new things at extremely fast rates [03:38:51] sure, but there is no point of storing algo results - i would love for an algo to suggest missing data and find incorrect values. But "storing" it implies level of validity. If you save algo result, it is no different than your browser caching a web site [03:39:09] yurik, the idea is different from the google knowledge graph, though, because, the nodes in the google knowledge graph are not machine parsable as "facts" where the knowledge graph i am invisioning would be [03:39:28] envisioning [03:39:37] forget about google - any algorithm can generate data - why store that result? [03:39:49] you usually store algo results ONLY to optimize performance [03:39:56] so that you don't have to re-run the algorithm [03:40:29] yurik, store "facts" so the algorithm knows what questions to ask, what blanks need to be filled [03:40:34] yurik, am i making any sense? [03:41:37] i must be missing your point - are you talking about an algo that generates more wikidata data, or to reason things out based on wikidata ? [03:42:04] yurik, sure, wikidata data. [03:42:12] yurik, im really talking about any knowledge base. [03:42:28] yurik, wikidata just so happens to be the largest knowledge base that i know of [03:43:03] again - algo to generate more data for the knowledge base, or an algo that uses existing knowledgebase to answer questions? [03:43:51] yurik, algo that uses existing data to ASK questions [03:44:06] what's the goal of asking questions? [03:44:07] yurik, then algo to parse internet data to answer questions and fill knowledgebase [03:44:18] what other goal would there be dude [03:44:23] so two algos that talk to each other? :D [03:44:27] oh my god [03:44:50] one algo to ask questions using existing data [03:45:06] the other algo to find data that answers the question, fills knowledgebase with new data [03:45:13] question answered [03:45:29] algo finds more missing data, asks another question yurik [03:45:34] yurik, im trying to build fucking skynet [03:45:39] good luck [03:45:47] lol [03:45:57] sorry im so serious [03:46:06] everything in my life is serious as fuck [03:46:16] yurik, do you understand me now though? [03:46:23] not really ) [03:46:26] yurik, are you a software developer? [03:46:39] no, i am the architect of the existing skynet [03:46:44] -.- [03:46:49] well [03:46:56] that would make you a programmer then wouldnt it? [03:47:03] guess so [03:47:07] dude [03:47:47] you study philosophy yurik [03:47:54] you should know what im getting at [03:48:04] yurik, ive been thinking about what knowledge ACTUALLY IS [03:48:19] yurik, is subject-predicate statements with varying degrees of certainty [03:48:23] yurik, do you get me [03:48:38] yurik, these statements are interrelated [03:48:54] yurik, they are expressed as objects which have properties [03:49:48] yurik, apple is red, apple is spherical , apple = { color: red, shape : spherical } [03:51:51] yurik, oh you dont study philosophy [03:54:37] not really. I am very much grounded. My point above was this: if you have some algo (it doesn't matter if that algo has two parts - one asking questions and the other answering it), it can produce any kind of data one may need. If you look at Wikidata, the idea is that behind each piece of data, you have a human that produced it. Possibly using an algo, but still putting some thoughts into it. It is possible to create other systems like wolfram [03:54:37] alpha - that will create any kind of reasoned-out dataset. But there is no point in persisting it together with the human curated data - they are simply too different in nature [03:55:08] everything beyond this is philosophy, and i doubt i am well versed on the deeper subject [04:02:52] yurik, i thought wikidata might serve as a good base knowledge base. [04:03:05] maybe not [04:05:22] base - sure. As long as the algo doesn't attempt to modify it [04:05:49] yurik, i am creating my own knowledgebase [04:06:08] yurik, i am just here because it is the only knowledge base with a channel [04:06:16] and subject-matter experts [04:06:21] yurik, thanks again for your help [04:06:45] sure thing, and good luck! crazy ideas sometimes result in amazing results ) [04:07:13] yurik, prepare for skynet [04:07:55] yurik, have you heard of lojban? [10:01:18] Are you around, sjoerddebruin? [10:02:40] A bit [10:03:07] :D [10:03:40] Perhaps Q29 should be semiprotected again (indefinitely?) [10:03:59] https://www.wikidata.org/w/index.php?title=Special:Log&page=Q29 [10:04:39] done [10:05:29] Thanks! :) [17:34:24] good evening