Post AbkOCB4STRfc2rFpq4 by Silicanus@the.asbestos.cafe
 (DIR) More posts by Silicanus@the.asbestos.cafe
 (DIR) Post #AbbMA5cF7eT60sRHHM by julia@eepy.moe
       2023-11-08T15:55:37.349Z
       
       4 likes, 4 repeats
       
       I am #begging people to #stop #writing their posts like #this. It doesn't help #discoverability, it just harms #readability.
       
 (DIR) Post #AbbMCUFMW3hOz3sbnk by adiz@soc0.outrnat.nl
       2023-11-08T16:03:32.601Z
       
       1 likes, 0 repeats
       
       @julia@eepy.moe Mastodon problem since they're all Twitter-brain[dead].
       
 (DIR) Post #AbbMvkCu1T0XW0JGXQ by theorytoe@ak.kyaruc.moe
       2023-11-08T16:11:46.444270Z
       
       0 likes, 1 repeats
       
       @julia #let #me #help #you #with #that #one#this #should #be #a #better #way #to #write #posts
       
 (DIR) Post #AbbN2O3wkmeArXHvmK by tomo@fedi.azumanga.gay
       2023-11-08T15:56:21.401908Z
       
       2 likes, 0 repeats
       
       @julia #ok #but #it #is #kinda #funny
       
 (DIR) Post #AbbN3JdigS89QzSnx2 by MercurialBuilding@cawfee.club
       2023-11-08T16:13:07.216254Z
       
       1 likes, 2 repeats
       
       @julia #kid #named #finger
       
 (DIR) Post #AbbPvJE6dT2XhOMBCy by lizzie@brain.worm.pink
       2023-11-08T16:45:18.283231Z
       
       0 likes, 0 repeats
       
       @julia inline tags were a mistake
       
 (DIR) Post #AbbUX4ENElNy0lk3rU by lamp@kitty.haus
       2023-11-08T17:36:56.324472Z
       
       0 likes, 0 repeats
       
       @julia @enigmatico Lolkevin karhan moment
       
 (DIR) Post #AbbeWVxzd6SJHhSHQ0 by pry@raru.re
       2023-11-08T19:28:36Z
       
       0 likes, 0 repeats
       
       @julia i also just hate tagging systems. personally i find that it really doesnt help with anything
       
 (DIR) Post #AbbmNv2np0u65X7Jrc by scathach@stereophonic.space
       2023-11-08T20:56:48.272183Z
       
       0 likes, 0 repeats
       
       @pry @julia What about tagging systems do you find unhelpful?
       
 (DIR) Post #AbbmT8Fk8iAyVCeiNU by meso@the.asbestos.cafe
       2023-11-08T20:57:55.269091Z
       
       0 likes, 0 repeats
       
       @julia it doesnt "harm readability" its just annoying and retarded
       
 (DIR) Post #AbbmWz1UOjmf2CpCqG by kirby@lab.nyanide.com
       2023-11-08T20:58:35.439434Z
       
       0 likes, 0 repeats
       
       @julia new mrf idea cc @mint plz plz plz
       
 (DIR) Post #AbbmcCcBI83SdXTZNQ by sun_eater@hidamari.apartments
       2023-11-08T20:59:32.962651Z
       
       0 likes, 1 repeats
       
       @meso @julia are you telling me it doesn't hurt you when you read it?
       
 (DIR) Post #AbbmfAzFnfqASNgasy by meso@the.asbestos.cafe
       2023-11-08T21:00:06.563235Z
       
       0 likes, 0 repeats
       
       @sun_eater @julia No i dont care i read it and how stupid it is takes priority
       
 (DIR) Post #AbbmhrNn926QWYRXBA by mint@ryona.agency
       2023-11-08T21:00:04.165073Z
       
       1 likes, 1 repeats
       
       @kirby @julia Might be too much effort for small return. Apparently, tags are in object metadata, so aside from replacing words you'd also have to add it to their field.
       
 (DIR) Post #AbbmjpwlSEZGeLkBd2 by kirby@lab.nyanide.com
       2023-11-08T21:00:48.664408Z
       
       0 likes, 0 repeats
       
       @mint @julia FUCK
       
 (DIR) Post #AbbmouvwLE4lZ0mTZo by sun_eater@hidamari.apartments
       2023-11-08T21:01:51.114746Z
       
       0 likes, 0 repeats
       
       @meso @julia masochist
       
 (DIR) Post #AbbmrFHXeouxFhzlEu by meso@the.asbestos.cafe
       2023-11-08T21:02:17.446716Z
       
       0 likes, 0 repeats
       
       @sun_eater @julia its pleasure to laugh at it instead of even considering the person's point
       
 (DIR) Post #Abbmtp01pbygbj229g by meso@the.asbestos.cafe
       2023-11-08T21:02:45.121689Z
       
       0 likes, 0 repeats
       
       @sun_eater @julia This is what people called cringe. I think people should socially shame things as cringe more. Nooo you cant call it cringe we should be accepting Nah dawg youre CRINGE kys
       
 (DIR) Post #Abbn8pqIykNrZAdukq by sun_eater@hidamari.apartments
       2023-11-08T21:05:25.755212Z
       
       0 likes, 1 repeats
       
       @meso @julia it amazes me people still use hashtags tbhIt was already cringe back when it started
       
 (DIR) Post #AbbnHc4dXZ5DT9X2cS by sun_eater@hidamari.apartments
       2023-11-08T21:06:58.613900Z
       
       0 likes, 0 repeats
       
       @meso @julia if at least people used it like those instagram accounts that makes the post, then put the hashtags at the very bottom of it so it won't show unless people click read more
       
 (DIR) Post #AbbuUtpbWqvZzFmcM4 by pry@raru.re
       2023-11-08T22:27:49Z
       
       1 likes, 0 repeats
       
       @scathach @julia with multiple people, coordinating tags is extremely hard. Tim Berners-Lee actually has a good point about why hypertext knowledge systems and knowledge organization systems which many people need to contribute to should not use tagging.
       
 (DIR) Post #Abbvg7pvtimMYyqzB2 by scathach@stereophonic.space
       2023-11-08T22:40:59.520168Z
       
       0 likes, 0 repeats
       
       @pry @julia There's a really interesting article I read last summer that made the opposite argument using emergent folksonomies on fanfiction sites as a case study, I think this was it though I'm not certain https://thehistoryoftheweb.com/postscript/fandom-folksonomies-and-tagging-things-right/
       
 (DIR) Post #Abc5iFTxRS0aOdE53o by kkarhan@mstdn.social
       2023-11-08T22:55:17Z
       
       0 likes, 0 repeats
       
       @julia wouldn't it he better if :activitypub: clients were to properly handle #hashtags and if #discoverability wasn't useless because making it a hidden #OptIn setting means noone I know ever activated it successfully...I think :mastodon: and other implementations should just have an option to display the # instead...Sadly using hashtags is a necessarx evil because #FullTextSearch - even one's own posts - DOES NOT WORK!!!
       
 (DIR) Post #Abc5iGEkdTRGjl7Rzs by lamp@kitty.haus
       2023-11-09T00:33:33.065918Z
       
       0 likes, 0 repeats
       
       @kkarhan @julia Doesn't mastodon have opt-in full text search now?
       
 (DIR) Post #AbcAobInjBNbZur7mC by pry@raru.re
       2023-11-09T01:30:20Z
       
       0 likes, 0 repeats
       
       @scathach @julia ooh lemme take a look,  that sounds interesting
       
 (DIR) Post #AbcJnLyo6syLM1z4YC by pry@raru.re
       2023-11-09T03:11:18Z
       
       0 likes, 0 repeats
       
       @scathach I think it points out what I tend to dislike about tagging:"Folksonomies are not without their problems. How does one, for instance, prevent a system where each tag is entirely unique and unfindable. What about misspellings? Or synonyms? Or words that have different meanings in different cultural contexts? And of course, even if you were able to smooth all of those issues, and dozens more, out, there’s still not a great way to incorporate the personal tags that enhance rather than categorize the content. How do we, in other words, account for #FedoraDad?"They later mention how Ao3 has a job specifically for people fixing inconsistencies and such of tags:"To bring a sortable, overarching structure to tags, AO3 created a volunteer group of individuals known as Tag Wranglers. The job of the Tag Wrangler is to try to find balance and even out the inconsistencies between tags. They fix misspellings, they combine synonyms, they make sure that tag groups remain meaningful. But a Tag Wrangler only seeks to better organize content so it can be found easier by fans. Creative asides, absurd misnomers, or personal references are actively left alone."I will need to read more about Curated Folksonomies but I think that these sorts of problems are too fundamental and I don't like the idea of hiring people to fix up these sorts of problems. For my own personal "knowledge system", I would much rather have mechanical automatic clustering which I can customize if I notice the machine has made an error.
       
 (DIR) Post #AbcJzk68GVADd9Jlaq by nemesis@the.asbestos.cafe
       2023-11-08T22:48:00.640099Z
       
       1 likes, 0 repeats
       
       @scathach @julia @pry the coordination issue is a good point against hierarchical ontologies, its not a good argument against tags I don't think.  c.f. https://web.archive.org/web/20050601013309/http://shirky.com/writings/ontology_overrated.htmlto the extent tags are similar to each other you can detect that by just noting that they correlate with each other, which gives a way to automatically solve the coordination problem
       
 (DIR) Post #AbcJzkn1h1TVmBO1S4 by pry@raru.re
       2023-11-09T03:13:32Z
       
       0 likes, 0 repeats
       
       @nemesis @julia @scathach I see. Maybe this is too ambitious or lofty but for the use case that I am thinking of (organizing documents of knowledge), I would much rather have machines plot the documents in some high dimensional space and then cluster documents based on this "semantic vectorization"
       
 (DIR) Post #AbcNMTGXJRnlwuiIyG by scathach@stereophonic.space
       2023-11-09T03:51:11.224429Z
       
       0 likes, 0 repeats
       
       @pry That makes sense.  What do you mean by mechanical automatic clustering?>What about misspellings? Or synonyms? Or words that have different meanings in different cultural contexts?For Leibowitz, the (currently) single-user booru/wiki I'm working on, I plan to solve these using some combination of Levenshtein distance, linguistic analysis provided by a search engine library, and/or machine learning
       
 (DIR) Post #AbcNZfjuFNKuOskJJg by scathach@stereophonic.space
       2023-11-09T03:53:35.369484Z
       
       0 likes, 0 repeats
       
       @pry @nemesis @julia Oooh is that what you mean by mechanical automatic clustering?  I'd like to do something like this with Leibowitz as well to link semantically similar documents, though it will probably be orthogonal to the existing hierarchical tag system.
       
 (DIR) Post #AbcNmcgotgzohqTgUS by nemesis@the.asbestos.cafe
       2023-11-09T03:53:11.897395Z
       
       1 likes, 0 repeats
       
       @scathach @pry compute vector text embeddings (with an LLM, or even just with word2vec + bag of words) and then use k-means clustering or something like it (e.g. GMM)
       
 (DIR) Post #AbcNmfDrTocAYtx6Gm by nemesis@the.asbestos.cafe
       2023-11-09T03:53:38.487993Z
       
       0 likes, 0 repeats
       
       @scathach @pry maybe add a PCA step in between those
       
 (DIR) Post #AbcNnx1YmOJE3uoPCa by pry@raru.re
       2023-11-09T03:56:11Z
       
       1 likes, 0 repeats
       
       @nemesis @scathach yes exactly. word embeddings are surprisingly good ive found. and I think sentence transformers is probably what ill use for my personal project. additionally im been wanting to experiment with vector databases which are build to index and cluster vector data. a lot of them will use Hierarchical Navigable Small Worlds (HNSW) as an algorithm for this.
       
 (DIR) Post #AbcOAot5CK4xUFkerA by pry@raru.re
       2023-11-09T04:00:18Z
       
       1 likes, 0 repeats
       
       @scathach yes exactly. id love to know more about what you wanna do with Leibowitiz.im going to spend my week long break working on a notetaking system that automatically uses semantic embeddings to cluster notes together as well as summarize documents. im trying to come up with a decent ui as well as find a good way of interacting with the models as opposed to simply accepting them as they decide organizations. also i think there is a tension between automatic clustering and manual cataloging where doing it manually helps retention. personally i think there is a good middle ground and i would like to use something like this to automate clustering my notes.
       
 (DIR) Post #AbcOIbIFckbhdtEZ0K by nemesis@the.asbestos.cafe
       2023-11-09T03:57:14.928838Z
       
       0 likes, 0 repeats
       
       @pry @scathach somehow I have failed to hear of this algorithm, thank you for mentioning it
       
 (DIR) Post #AbcOIc8MV0I6FVcBEG by pry@raru.re
       2023-11-09T04:01:38Z
       
       0 likes, 0 repeats
       
       @nemesis @scathach totally! you can find more stuff if you look up "approximate nearest neighbor algorithms"funny how i get interested in this stuff just as hype around LLMs reaches insufferable levels. lots of money is being poured into vector databases for LLMs.
       
 (DIR) Post #AbcXjqTACuFS7GP88O by nemesis@the.asbestos.cafe
       2023-11-09T04:03:20.395889Z
       
       0 likes, 0 repeats
       
       @pry @scathach thanks.  I'm familiar with vector quantization, but not approximate nearest neighbor algorithms
       
 (DIR) Post #AbcXjrKgzt4AnHRsZM by pry@raru.re
       2023-11-09T05:47:29Z
       
       0 likes, 0 repeats
       
       @nemesis yea totally :)
       
 (DIR) Post #AbdH47jEFAfvNCxipk by 7666@comp.lain.la
       2023-11-09T14:15:26.449454Z
       
       0 likes, 0 repeats
       
       @pry @scathach @julia private trackers, boorus, and hentai galleries would like a word with you
       
 (DIR) Post #AbdHQBx1E2Y0q7w3Fo by pry@raru.re
       2023-11-09T14:19:23Z
       
       0 likes, 0 repeats
       
       @7666 @julia @scathach that's fair. in many cases its still extremely non-trivial to automate semantic vectorization. but I will say I imagine it takes a surprising amount of actual peoples time organize boorus and hentai galleries.
       
 (DIR) Post #AbdHZtwp125VtqTOWe by 7666@comp.lain.la
       2023-11-09T14:21:12.280846Z
       
       0 likes, 0 repeats
       
       @pry @julia @scathach It does take a lot of time, but users are properly incentivized to do so and for the benefit of the entire community. It's likely that the amount of time saved with proper tagging exceeds the up front cost by many factors.
       
 (DIR) Post #AbkOCB4STRfc2rFpq4 by Silicanus@the.asbestos.cafe
       2023-11-13T00:38:08.105774Z
       
       0 likes, 0 repeats
       
       @julia is #racemixing morally #permissible? please #discuss
       
 (DIR) Post #Ac0Vm1sYTdkFbRgjJo by vintprox@techhub.social
       2023-11-20T19:18:34Z
       
       0 likes, 0 repeats
       
       @pryThat's not entirely fair, as you could use #hashtags to show the topic(s) of intent. I think what @julia tried to point out is an excessive and purposeless insertion.