Papers of the day   All papers

Neural Decipherment via Minimum-Cost Flow: from Ugaritic to Linear B

Comments

Nando de Freitas: Machine learning to automatically translate long-lost languages https://www.technologyreview.com/s/613899/machine-learning-has-been-used-to-automatically-translate-long-lost-languages/

8 replies, 446 likes


Ben Sprecher: Holy shit, and of course. We now have the ultimate Rosetta Stone. NN for automatic decipherment of lost languages: - Ugaritic, +5.5% vs. SOTA. - the first automatic deciphering of Linear B (related to ancient Greek), correctly translates 67.3% https://arxiv.org/pdf/1906.06718.pdf?utm_source=Machine+Learning+Yearning&utm_campaign=af9c19872d-EMAIL_CAMPAIGN_2019_05_22_06_05_COPY_01&utm_medium=email&utm_term=0_bd0078304d-af9c19872d-115473869

6 replies, 172 likes


Peter Gainsford: So, this is making news: automated 'decipherment' of ancient languages, touting a 67.3% success rate with Linear B. Thing is, they're not deciphering languages. The project is much more limited: they take a SCRIPT and compare it to a known language. https://arxiv.org/abs/1906.06718 https://t.co/Mm1dHXAs5Q

3 replies, 56 likes


Ancient History Encyclopedia: Linear B Script was the #writing system used by the Mycenaean civilization. Examples of this script have been recovered from late Minoan II contexts in Crete and Mycenaean IIIA-B contexts in mainland Greece. http://ancient.eu/Linear_B_Script/ #History https://t.co/KtHUdnwssb

0 replies, 32 likes


Ben Cartlidge: We already have machines that can learn how to decipher Linear B. They're called "people". Some if not most of them work at things called "classics departments". Talk to them. They'll be able to translate any Linear B you're having issues with.

1 replies, 24 likes


MIT Technology Review: Researchers have developed a machine-learning system capable of deciphering lost languages. https://trib.al/t7mlRYZ

0 replies, 16 likes


Jens Notroff: Ah, the burden of bold headlines again (or subheadlines in this case) ... Important notes regarding recent tech news about the automated 'deciphering' of dead languages in this short thread (spoiler: it's only as good as the data basis accumulated by human research 🤷‍♂️).

0 replies, 5 likes


Ben Cartlidge: Sob/lol/scream Tech people, PLEASE STOP DOING THIS.

0 replies, 5 likes


Simon de la Rouviere: Was waiting for this to happen! Machine learning to decipher patterns in lost languages. 😍 Imagine deciphering all this? At least pushing the needle? https://en.m.wikipedia.org/wiki/Undeciphered_writing_systems

0 replies, 2 likes


Hacker News: Deciphering Linear B with AI https://arxiv.org/abs/1906.06718

1 replies, 2 likes


Gordon Brander: 🐳 this, but for whales. “We propose a novel neural approach for automatic decipherment of lost languages.” https://arxiv.org/abs/1906.06718

1 replies, 2 likes


Therfer: #Linguistics and #MachineLearning Work in https://arxiv.org/abs/1906.06718 https://trib.al/PuNI0hq

0 replies, 2 likes


space octopus: this is actually pretty cool, even if you don't know what it's about. trust me. it's cool.

0 replies, 1 likes


IntuitionMachine: Really cool research using neural networks to decipher 'lost human languages': https://arxiv.org/abs/1906.06718 #deeplearning #ai

0 replies, 1 likes


Robert (Munro) Monarch: Interesting new paper applying Machine Translation to Linear B. https://arxiv.org/abs/1906.06718 But is it really a "lost" language? It's Ancient Greek and the script was deciphered 50+ years ago. Maybe @linearb has an opinion?

0 replies, 1 likes


Seth Largo: "In this paper we propose a novel neural approach for automatic decipherment of lost languages." Mode correctly translates Linear B cognates at 67%. https://arxiv.org/pdf/1906.06718.pdf?utm_source=Machine+Learning+Yearning&utm_campaign=af9c19872d-EMAIL_CAMPAIGN_2019_05_22_06_05_COPY_01&utm_medium=email&utm_term=0_bd0078304d-af9c19872d-115473869 h/t @dpinsen

0 replies, 1 likes


rogueclassicist: Thread 🔻🔻 https://twitter.com/PeterGainsford/status/1146224111535812610

0 replies, 1 likes


Poornima Farrar: Leveraging #unsupervised #machinelearning for deciphering lost languages: https://arxiv.org/abs/1906.06718

0 replies, 1 likes


Content

Found on Jul 01 2019 at https://arxiv.org/pdf/1906.06718.pdf

PDF content of a computer science paper: Neural Decipherment via Minimum-Cost Flow: from Ugaritic to Linear B