Papers by Annette Hautli-Janisz
lingvis.io - A Linguistic Visual Analytics Framework (P19-3)
Copied to clipboard
Mennatallah El-Assady, Wolfgang Jentner, Fabian Sperrle, Rita Sevastjanova, Annette Hautli-Janisz, Miriam Butt, Daniel Keim
| Challenge: | Using a modular framework, linguistic visual analytics applications can be rapidly prototypized using a web-based framework. |
| Approach: | They propose a modular framework for rapid prototyping of linguistic, web-based, visual analytics applications. |
| Outcome: | The proposed framework supports rapid prototyping of linguistic, web-based, visual analytics applications. |
Probing Bias Formation in Medical LLMs through Activation Steering (2026.acl-srw)
Copied to clipboard
| Challenge: | Large Language Models specialized for the medical domain achieve high performance on static benchmarks, but are vulnerable to sycophantic confabulation. |
| Approach: | They propose a framework toward clinical AI systems that are more robust and aligned with expert medical logic. |
| Outcome: | The proposed framework outperforms static global interventions on a medical prompt with cluster-conditioned dynamic steering. |
PSE v1.0: The First Open Access Corpus of Public Service Encounters (2024.lrec-main)
Copied to clipboard
Ingrid Espinoza, Steffen Frenzel, Laurin Friedrich, Wassiliki Siskou, Steffen Eckhard, Annette Hautli-Janisz
| Challenge: | a dataset of public service encounters in germany provides a new research directive . data from the public service encounters are used to investigate bias, bureaucratic discrimination and other power-driven dynamics in the actual communication . |
| Approach: | They propose to compile a dataset of transcribed public service encounters in germany . they propose to open up the black box of direct state-citizen interaction . |
| Outcome: | The proposed dataset allows the community to open up the black box of direct state-citizen interaction. |
QT30: A Corpus of Argument and Conflict in Broadcast Debate (2022.lrec-1)
Copied to clipboard
| Challenge: | Broadcast political debate is the public's easiest access to opinions that shape policies and enables the general public to make informed choices. |
| Approach: | They present the largest corpus of analysed dialogical argumentation ever created using 30 episodes of BBC's 'Question Time' from 2020 and 2021. |
| Outcome: | The resource is freely available at http://corpora.aifdb.org/qt30. |
A Multilingual Approach to Question Classification (L18-1)
Copied to clipboard
| Challenge: | Existing work on questions has focused on understanding the structure of questions per se . a few approaches explicitly focus on information-seeking questions, but this work is either based on big data or crowdsourcing. |
| Approach: | They propose a dependency-parsed, parallel multilingual corpus of information-seeking and non-information-seeing questions . they employ a linguistically motivated rule-based system that uses linguistic cues from one language to help classify questions across other languages. |
| Outcome: | The proposed system correctly classifies questions in 79% of cases, compared to other systems. |