Dominic Petrak

Doctoral Researcher @UKPLab, @CS_TUDarmstadt | Working on improving the social acceptance of dialogue systems | Supervised by @NafiseSadat and @IGurevych

[Dominic Petrak @PetrakDominic]

Do you want to know what really improves task completion and factual consistency in task-oriented document-grounded dialogues?😏Then you might want to check out our latest preprint on arxiv! 🤓👉📰 https://t.co/sbIAv7Z88l
(1/🧵)

[UKP Lab @UKPLab]

𝗠𝗲𝗲𝘁 𝗙𝗘𝗗𝗜 💬🤖
Want to know what really improves task completion and factual consistency in task-oriented document-grounded dialogues? 😉

Then check out our #EMNLP2024 paper! 🤓

👉📰 https://t.co/i3Loc16lpb (1/🧵)

[Nico Daheim @ndaheim_]

Please drop by our poster tomorrow and talk to my co-authors about how IVON can directly optimize a variational objective with similar performance and cost as Adam but many things on top (uncertainty, model adaptation, diagnosis, ...)

Thu. 13:30-15:00, hall c 4-9, poster #1402

[Dominic Petrak @PetrakDominic]

Interested ⁉️🤓We provide FEDI and its generation framework on GitHub. 👩‍💻👉 https://t.co/UIGfnEtUFn
(9/ 🧵)

[Dominic Petrak @PetrakDominic]

Do you want to know what really improves task completion and factual consistency in task-oriented document-grounded dialogues?😏Then you might want to check out our latest preprint on arxiv! 🤓👉📰 https://t.co/sbIAv7Z88l
(1/🧵)

[Dominic Petrak @PetrakDominic]

FEDI consists of 8.8k dialogues generated and annotated using GPT-3.5-Turbo, 6k of which with annotations for implicit user feedback. For testing, we additionally provide a set of dialogues collected by humans.
(8/ 🧵)

[Nafise Sadat Moosavi @NafiseSadat]

Fully funded PhD position @shefcompsci, @SheffieldNLP & Healthy Lifespan Institute
Join us in developing cutting-edge NLP techniques for inclusive & evidence-based recommendations to enhance dietary guidelines for older adults
Deadline: February 01
https://t.co/8QYkEsaQUq

[Dominic Petrak @PetrakDominic]

This is me presenting our work on free-text human feedback at #EMNLP2023. I really enjoyed over two hours of discussions on potential future work and all the positive feedback. Very encouraging!😊
@UKPLab @sheffielduni @NafiseSadat @wluper_ @ai_nikolai @CS_TUDarmstadt @IGurevych

[Dominic Petrak @PetrakDominic]

Finally... 😊 Check out our EMNLP 2023 paper on learning from free-text human feedback now on arxiv (https://t.co/ETHtHqYEK6)!

We investigate the feasibility of augmenting existing dialogue datasets with the needed annotations and propose two new taxonomies for this purpose.

[UKP Lab @UKPLab]

Human feedback on poor dialog system performance is important for the development of #ConvAI. 👍👎

But datasets with free-text human feedback are scarce. Can data augmentation come to the rescue?

A 🧵 on one of our #EMNLP2023 papers (1/11).

📰 https://t.co/E3YEmk7gh7 #NLProc

[Sheffield NLP @SheffieldNLP]

📢 We're happy to announce that our group has 12 papers accepted to #EMNLP2023 (6 main, 6 findings) 🧵⬇️

Congratulations to all our members and collaborators! 🥳 #NLProc

[UKP Lab @UKPLab]

7 papers (co-)authored by UKP accepted as long main conference papers at @emnlpmeeting! Congratulations to everybody involved! 💐 (1/🧵) #EMNLP2023

[UKP Lab @UKPLab]

Two UKP papers have been accepted to *SEM 2023 (@_starsem) in July in Toronto. The preprint »Arithmetic-Based Pretraining« by @PetrakDominic, @NafiseSadat and @IGurevych has just been finalized – you can find the new, updated version here: https://t.co/0DBlVQD3it

[Dominic Petrak @PetrakDominic]

TL;DR: We propose a new, extended pretraining approach that uses contrastive learning and character-level representation to improve the capacity of pretrained language models to work with numbers. 🤓Check it out! It's simple and efficient. Code will be published asap!

[Dominic Petrak @PetrakDominic]

Interested in improving the numeracy of your pretrained Transformer-based model without retraining or architectural changes? Then check out the updated preprint of our paper "Arithmetic-Based Pretraining" (https://t.co/uCxrEAUWE2), which will also be presented at @_starsem! 🥳

[UKP Lab @UKPLab]

Two UKP papers have been accepted to *SEM 2023, taking place July 13/14 co-located w/ #ACL2023!
Read the pre-prints here:
https://t.co/ursdB3eEWX
https://t.co/MSTL2VkxvY
Congratulations to the authors @devnull90, @AndreasWaldis, @PetrakDominic, @NafiseSadat & @igurevych! #SEM2023