BamNLP

BamNLP

@bamnlp.de

BamNLP Research Group at the University of Bamberg, Germany We work on #nlproc #nlp #computationalpsychology #deeplearning #emotions #sentiment #argumentmining PI: @romanklinger.de Website: https://www.uni-bamberg.de/en/nlproc/#

394 followers133 following38 posts

BamNLP
@bamnlp.de

EMNLP in Budapest is around the corner, one of the flagship #NLPROC conferences. We are represented with 4 contributions. Come to talk to us, if you are there!

View full thread
BamNLP
@bamnlp.de

Christopher Bagdon presents our work on multimodal emotion recognition from social media: Christopher Bagdon, Carina Silberer, Roman Klinger: Emotion Experience, Expression, and Perception: Emotion Analysis on Multimodal Social Media Posts. (EMNLP 2026 Main) arxiv.org/abs/2609.18385

arxiv.orgEmotion Experience, Expression, and Perception: Emotion Analysis on Multimodal Social Media PostsEmotions are an essential aspect of human communication, particularly on social media, where authors frequently combine text and images to convey their emotions. Yet prior work on emotion analysis of ...
BamNLP
@bamnlp.de

EMNLP in Budapest is around the corner, one of the flagship #NLPROC conferences. We are represented with 4 contributions. Come to talk to us, if you are there!

BamNLP
@bamnlp.de

Aswathy Velutharambath talks about our findings that recognizing lies from text doesn't work: Aswathy Velutharambath, Kai Sassenberg, Roman Klinger: What If Deception Cannot Be Detected? A Cross-linguistic Study on the Limits of Deception Detection from Text (CL Journal) doi.org/10.1162/COLI...

doi.orgWhat If Deception Cannot Be Detected? A Cross-linguistic Study on the Limits of Deception Detection from TextAbstract. Can deception be detected solely from written text? Cues of deceptive communication are inherently subtle, and even more so in text-only communication. Yet, prior studies have reported considerable success in automatic deception detection. In this study, we hypothesize that such findings are largely driven by artifacts introduced during data collection and do not generalize beyond specific datasets. We revisit this foundational assumption by introducing a belief-based deception framework, which defines deception as a misalignment between an author’s claims and their true beliefs, irrespective of factual accuracy, allowing deception cues to be studied in isolation. Based on this framework, we construct three corpora—collectively referred to as DeFaBel—including a German-language corpus of deceptive and non-deceptive arguments and a multilingual version in German and English, each collected under different conditions to account for potential belief change and enable cross-linguistic analysis. Using these corpora, we evaluate commonly reported linguistic cues of deception. Across all three DeFaBel variants, we find that these cues exhibit negligible and statistically insignificant correlations with deception labels, contrary to prior work that treats such cues as reliable indicators. We further benchmark against other English deception datasets that follow similar data collection protocols. While some of these show statistically significant correlations, the effect sizes remain low and, critically, the set of predictive cues is inconsistent across datasets. We also evaluate deception detection using feature-based models, pre-trained language models, and instruction-tuned large language models. While some of these models perform well on established deception datasets, they consistently perform near chance on DeFaBel. Our findings challenge the core assumption that deception can be reliably inferred from linguistic cues and call for a rethinking how deception is studied and modeled in Natural Language Processing.
BamNLP
@bamnlp.de

Tomorrow, #lrec2026 starts (for us)! On Tuesday, we are coorganizers of the Social on+NLPSI workshop, on Saturday of the CAS workshop. In between, we have a set of papers across various sessions.

BamNLP@bamnlp.de

#LREC2026 #NLProc papers from @bamnlp.de @uni-bamberg.de ! See you in Palma!

BamNLP
@bamnlp.de

Interested in real world probability (estimates) and how LLMs and humans misinterpret them? Want to do a PhD in #NLProc? Please get in touch and apply in this awesome project lead by @seanpapay.bsky.social (as part of @bamnlp.de at @uni-bamberg.de)

seanpapay.bsky.social@seanpapay.bsky.social

We’re hiring! A 3-year PhD position at the Chair for Funamentals of NLP at Uni Bamberg to study how LLMs represent and reason about probabilistic world knowledge. Full details & application info: www.uni-bamberg.de/fileadmin/ab...

uni-bamberg.deuni-bamberg.de
BamNLP
@bamnlp.de

Thanks again for your visit, the presentation, and all the interesting discussions we had!

Barbara Plank@barbaraplank.bsky.social

I visited @uni-bamberg.de this week to give a Bamberg Center for AI lecture, and had the opportunity to meet many brilliant researchers, thanks for the invitation @bamnlp.de, @bamdial.bsky.social and BaCAI committee

BamNLP reposted

Roman Klinger
@romanklinger.de

Tomorrow I'll give my inaugural lecture @uni-bamberg.de – if you don't know what language technology #nlproc is, and want to know what @bamnlp.de works on, please feel invited! If you can't make it at 6pm to Weberei 5 in Bamberg ("Erba Campus"), send me a message and I'll share a streaming link.

BamNLP
@bamnlp.de

Happening now: Sean Papay presents his work on CRFs for distant interactions in X5, 223 :-). #xllm #nlproc #acl2025

BamNLP
@bamnlp.de

Now happening in X5, 95: Johannes Schäfer presenting our work on default demographics in LLMs. #nlproc #acl2025

BamNLP
@bamnlp.de

Christopher Bagdon now presents his work on multimodal social media #emotion corpus creation. hall X4, far away from the entrance, stand 375.

BamNLP
@bamnlp.de

We will be at @aclmeeting.bsky.social in #Vienna next week! with some main and workshop papers, and an affinity group meeting. If you want to talk to us, please have a look at the group's schedule. We are looking forward to talking to you, please don't hesitate to get in touch! #NLProc #NLP #ACL2025

BamNLP
@bamnlp.de

Last week, we've been on a retreat to talk about the big picture of our research, research ideas, and synergies between group members – and did some team building. @annefriedrich.bsky.social was our external guest and helped us out of a couple of local optima. Thank you for participating!

BamNLP
@bamnlp.de

Organized with involvement from our group 🤩 – thanks @aswathyve.bsky.social for taking the lead!

Aswathy Velutharambath@aswathyve.bsky.social

✨ The First NLPSI Workshop @ ICWSM 2025 is a wrap! Huge thanks to the speaker (Bennett Kleinberg), authors, reviewers, attendees, and co-organizers (Sofie Labat, Neele Falk, @florplaza.bsky.social, @romanklinger.de & Veronique Hoste) who made it a success. nlpsi-workshop.github.io/highlights/

BamNLP
@bamnlp.de

We are really happy about this new collaboration, in which we use our experience with #LLM models for automatic text simplification in #psychology research.

Leibniz-Institut für Psychologie (ZPID)@zpid.bsky.social

Large Language Models für allgemeinverständliche Zusammenfassungen psychologischer Übersichtsarbeiten: Kooperation des @zpid.bsky.social mit der @uni-bamberg.de & @bamnlp.de nlp.de #psychologie #llm #klarpsy leibniz-psychology.org/news/detail/...

leibniz-psychology.orgMit Hilfe von Large Language Models allgemeinverständliche Zusammenfassungen psychologischer Übersichtsarbeiten verfassenKooperation des ZPID mit der Universität Bamberg
BamNLP
@bamnlp.de

Are you at #COLING2025? Come visit our poster presented by Yarik Menchaca Resendiz on "MOPO: Multi-Objective Prompt Optimization for Affective Text Generation". (Fri, Jan 24 13:30-15:00) #NLP #nlproc #prompts #llm #emotions

BamNLP
@bamnlp.de

We got one paper accepted to #COLING2025: "MOPO: Multi-objective prompt optimization for affective text generation." with @romanklinger.de and Yarik Menchaca Resendiz. We optimize prompts for multiple objectives, while indirectly optimizing the prompts that optimize the prompts #NLP #NLProc (1/3)

View full thread
BamNLP
@bamnlp.de

Have a look at our new #preprint on our work on human-in-the-loop prompt optimization. It's a preprint of a demo paper describing our experiment environment that we use in the @dfgpublic.bsky.social funded project INPROMPT. #nlp #nlproc

arxiv cs.CL@arxiv-cs-cl.bsky.social

Jiahui Li, Roman Klinger iPrOp: Interactive Prompt Optimization for Large Language Models with a Human in the Loop https://arxiv.org/abs/2412.12644

BamNLP
@bamnlp.de

New #preprint by @nlp-amelie.bsky.social and @romanklinger.de on automatically optimizing a claim paraphrased/extractor with preference optimization, such that the claims can be better fact-checked. #nlp #nlproc

arxiv cs.CL@arxiv-cs-cl.bsky.social

Amelie W\"uhrl, Roman Klinger Self-Adaptive Paraphrasing and Preference Learning for Improved Claim Verifiability https://arxiv.org/abs/2412.11653

BamNLP
@bamnlp.de

We got one paper accepted to #COLING2025: "MOPO: Multi-objective prompt optimization for affective text generation." with @romanklinger.de and Yarik Menchaca Resendiz. We optimize prompts for multiple objectives, while indirectly optimizing the prompts that optimize the prompts #NLP #NLProc (1/3)

BamNLP
@bamnlp.de

This leads to more robust prompts, because we can, a posteriori, decide for which objective a prompt should work, without retraining. We focus our evaluation on affective text generation across multiple domains, but it should work across other multi-objective use cases (2/3).

BamNLP
@bamnlp.de

Today, Eyke Hüllermeier gave a talk in our BACAI lecture series, on uncertainty quantification in machine learning. The main message was the difference between inherent randomness and a lack of knowledge: aleatoric and epistemic uncertainty.

BamNLP
@bamnlp.de

Hello Everybody! We are the Bamberg NLP group at the University of Bamberg. We exist since March 2024 and work on the interaction of #computerscience #naturallanguageprocessing #nlproc #psychology #computationalsocialsciences #digitalhumanities. If you want to learn more: www.uni-bamberg.de/nlproc/

You are up to date.