The below is an off-site archive of all tweets posted by @texttheater ever

December 4th, 2021

French army band medleys Daft Punk following Bastille Day parade youtu.be/vhQkku7jLfo via @YouTube

via Twitter Web App

Konzentrierte Vollpflege vs. unkonzentrierter Vollhorst

via Twitter Web App

Einreichungen in Computer Modern, wenn die Vorlage Times vorgibt. Weiß man da schon mehr? Wie passiert das? Warum machen Menschen das?

via Twitter Web App

RT @c_drosten: Ein sehr lesenswerter Beitrag von Georg Mascolo in der Süddeutschen Zeitung zur Rolle der Medien in der Pandemie (leider nic…

via Twitter for Android

RT @labenzticker: Sprakensehl, die: Die Sprache, die die Google-Buchdigitalisierungssoftware aus alten Frakturschriften extrahiert und die…

via Twitter for Android

RT @Defensie: @HengelaarGert @MinPres Beste Gert, dank voor je belangstelling in de @kon_luchtmacht! Inderdaad, vliegen lijkt magisch, maar…

via Twitter for Android

RT @jae_strandberg: Interesting portmanteau of glögg/glögi (Nordic mulled wine) and Möet champagne in the name of this Finnish "bubbling gl…

via Twitter for Android

@zehavoc @CChiarcos Would be cool to reimplement it like you suggest, but atm it's call custom Python code 🙈

via Twitter for Android

@PhilHackemann "Wer geimpft / genesen UND getestet ist, stellt nachweislich keine Gefahr dar." Das stimmt nicht. Warum schreiben Sie so einen Quatsch.

via Twitter Web App in reply to PhilHackemann

There is prior art on converting from UD to RRG by @CChiarcos and Christian Fäth, with which we perform an in-depth qualitative comparison.

I will post the link to our paper here soon.

via Twitter Web App in reply to texttheater

How did we measure this? Not satisifed with existing tree-distance measures, we defined a new one, which explicitly tries to predict which nodes annotators will relabel, delete, insert, or reattach. The measure is called Bottom-up Replugging (BURP). https://t.co/DuwY4utGw7 https://t.co/fSbeCJFFFi

via Twitter Web App

The converted trees serve as the starting point for manual annotation. Once we have collected around 2000 sentences, we can train a statistical parser that outperforms the rule-based conversion and serves as a better starting point for subsequent annotation. pic.twitter.com/in7PduARjQ

via Twitter Web App in reply to texttheater

For this purpose, we formalize RRG as a tree grammar whose derivation trees are isomorphic to UD trees. Operations correspond to RRG's juncture-nexus types. We use a bunch of rules to relabel UD trees into RRG derivation trees. pic.twitter.com/tDUrugHpyp

via Twitter Web App in reply to texttheater

In our upcoming @ud_workshop paper, Tatiana Bladier, Laura Kallmeyer, @Simon_Petitjean and I describe how we bootstrapped RRGparbank using UD parsing and rule-based conversion from Universal Dependencies to RRG. pic.twitter.com/hvOAi3PhdA

via Twitter Web App in reply to texttheater

RRGparbank is a new Role and Reference Grammar (RRG) parallel treebank based on George Orwell’s novel 1984. rrgparbank.phil.hhu.de pic.twitter.com/HM8WlOQrRU

via Twitter Web App

RT @tommdidomm: Dies getippt: "#wasmachteigentlichderchef 2021-12-03"
"U-Bahn. Seit zwei Wochen nich"
https://t.co/iznEHahwQU
Freundlichst…

via Twitter for Android

brissle obsessed with this insane photo from a honey wholesale website pic.twitter.com/2ONmVmNMsi

via Twitter for iPhone (retweeted on 7:22 AM, Dec 4th, 2021 via Twitter for Android)