@inproceedings{weber2026_116541,
  title = {Sounds Queer : Representation of LGBTQIA Identities in AI-generated Songs},
  author = {Weber, Sabine and McLeod, Andrew},
  year = {2026},
  pages = {4375-4394},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-116541x},
  url = {https://fis.uni-bamberg.de/handle/uniba/116541},
  language = {eng},
  abstract = {The expansion of generative artificial intelligence into the domain of music production has introduced new possibilities and risks for cultural representation, particularly in its depiction of marginalized groups. In this paper we study how LGBTQIA identities are represented in text-prompt-based end-to-end generated songs by the black-box models of Suno and Mureka. To do so we use a template-based approach to generate prompts containing identity terms like “gay”, “lesbian” or “genderqueer” that we pass to the examined systems, generating a dataset of 9600 songs. Using qualitative and quantitative content analysis of the generated song lyrics and musicological analysis of the generated music, we find that while songs generally convey high regard and positive sentiment towards queer individuals, they often reproduce stereotypes. We uncover representational harms specifically in the representation of aromantic, asexual, lesbian and intersex people. Musically, the models produce songs that strongly align with the requested genre, but show only minor differences based on the identity terms used in the prompts. We release the created Sounds Queer dataset and all associated code under open Creative Commons licensing for further research.},
  keywords = {Queer AI, NLP, Language Generation, Music Generation, Large Language Models, Stereotypes}
}

@inproceedings{weber2026_116538,
  title = {Sounds Queer : Representation of LGBTQIA Identities in AI-generated Songs},
  author = {Weber, Sabine and McLeod, Andrew},
  year = {2026},
  booktitle = {Proceedings of the 2026 ACM Conference on Fairness, Accountability, and Transparency},
  pages = {4375-4394},
  publisher = {ACM},
  address = {New York},
  isbn = {979-8-4007-2596-8},
  doi = {10.1145/3805689.3806752},
  url = {https://fis.uni-bamberg.de/handle/uniba/116538},
  language = {eng},
  abstract = {The expansion of generative artificial intelligence into the domain of music production has introduced new possibilities and risks for cultural representation, particularly in its depiction of marginalized groups. In this paper we study how LGBTQIA identities are represented in text-prompt-based end-to-end generated songs by the black-box models of Suno and Mureka. To do so we use a template-based approach to generate prompts containing identity terms like “gay”, “lesbian” or “genderqueer” that we pass to the examined systems, generating a dataset of 9600 songs. Using qualitative and quantitative content analysis of the generated song lyrics and musicological analysis of the generated music, we find that while songs generally convey high regard and positive sentiment towards queer individuals, they often reproduce stereotypes. We uncover representational harms specifically in the representation of aromantic, asexual, lesbian and intersex people. Musically, the models produce songs that strongly align with the requested genre, but show only minor differences based on the identity terms used in the prompts. We release the created Sounds Queer dataset and all associated code under open Creative Commons licensing for further research.},
  keywords = {Queer AI, NLP, Language Generation, Music Generation, Large Language Models, Stereotypes}
}

@inproceedings{greschner2026_115018,
  title = {Trust Me, I Can Convince You: The Contextualized Argument Appraisal Framework and the ContArgA Corpus},
  author = {Greschner, Lynn and Weber, Sabine and Klinger, Roman},
  year = {2026},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
  pages = {8327–8346},
  publisher = {European Languages Resources Association (ELRA)},
  address = {Paris},
  isbn = {978-2-493814-49-4},
  doi = {10.63317/484rpnvebop5},
  url = {https://fis.uni-bamberg.de/handle/uniba/115018},
  language = {eng}
}

@inproceedings{greschner2026_115020,
  title = {Categorical Emotions or Appraisals: Which Emotion Model Explains Argument Convincingness Better?},
  author = {Greschner, Lynn and Bauer, Meike and Weber, Sabine and Klinger, Roman},
  year = {2026},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
  pages = {8190–8203},
  publisher = {European Languages Resources Association (ELRA)},
  address = {Paris},
  isbn = {978-2-493814-49-4},
  doi = {10.63317/3vrvrgvtnvhn},
  url = {https://fis.uni-bamberg.de/handle/uniba/115020},
  language = {eng}
}

@inproceedings{weber2026_115023,
  title = {Less Is More?: The Role of Demographic Author Information in Emotion Classification of Ambiguous Text},
  author = {Weber, Sabine and Greschner, Lynn and Klinger, Roman},
  year = {2026},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
  pages = {8147–8161},
  publisher = {European Languages Resources Association (ELRA)},
  address = {Paris},
  isbn = {978-2-493814-49-4},
  doi = {10.63317/2cw8tpo82h55},
  url = {https://fis.uni-bamberg.de/handle/uniba/115023},
  language = {eng}
}

@inproceedings{weber2026_114414,
  title = {Says Who? : Argument Convincingness and Reader Stance Are Correlated with Perceived Author Personality},
  author = {Weber, Sabine and Greschner, Lynn and Klinger, Roman},
  year = {2026},
  booktitle = {The Proceedings for the 15th Workshop on Computational Approaches to Subjectivity, Sentiment Social Media Analysis (WASSA 2026)},
  pages = {265–277},
  publisher = {Association for Computational Linguistics},
  isbn = {979-8-89176-378-4},
  url = {https://fis.uni-bamberg.de/handle/uniba/114414},
  language = {eng}
}

@inproceedings{schafer2026_114413,
  title = {Which Demographics do LLMs Default to During Annotation?},
  author = {Schäfer, Johannes and Combs, Aidan and Bagdon, Christopher Doyle and Li, Jiahui and Probol, Nadine and Greschner, Lynn and Papay, Sean and Menchaca Resendiz, Yarik and Velutharambath, Aswathy and Wuehrl, Amelie and Weber, Sabine and Klinger, Roman},
  year = {2026},
  pages = {17331–17348},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-114413x},
  url = {https://fis.uni-bamberg.de/handle/uniba/114413},
  language = {eng},
  abstract = {Demographics and cultural background of annotators influence the labels they assign in text annotation – for instance, an elderly woman might find it offensive to read a message addressed to a “bro”, but a male teenager might find it appropriate. It is therefore important to acknowledge label variations to not underrepresent members of a society. Two research directions developed out of this observation in the context of using large language models (LLM) for data annotations, namely (1) studying biases and inherent knowledge of LLMs and (2) injecting diversity in the output by manipulating the prompt with demographic information. We combine these two strands of research and ask the question to which demographics an LLM resorts when no demographics is given. To answer this question, we evaluate which attributes of human annotators LLMs inherently mimic. Furthermore, we compare non-demographic conditioned prompts and placebo-conditioned prompts (e.g., “you are an annotator who lives in house number 5”) to demographics-conditioned prompts (“You are a 45 year old man and an expert on politeness annotation. How do you rate \{instance\}”). We study these questions for politeness and offensiveness annotations on the POPQUORN data set, a corpus created in a controlled manner to investigate human label variations based on demographics which has not been used for LLM-based analyses so far. We observe notable influences related to gender, race, and age in demographic prompting, which contrasts with previous studies that found no such effects},
  keywords = {-}
}

@article{weber2025_116845,
  title = {Can We Build AI That Does Not Harm Queer People?},
  author = {Weber, Sabine},
  year = {2025},
  journal = {Communications of the ACM},
  volume = {68},
  number = {5},
  pages = {21-23},
  publisher = {Association for Computing Machinery (ACM)},
  address = {New York, NY},
  issn = {1557-7317},
  doi = {10.1145/3720537},
  url = {https://fis.uni-bamberg.de/handle/uniba/116845},
  language = {eng}
}

@inproceedings{li2025_110866,
  title = {Language Models Are Poor Learners of Directional Inference},
  author = {Li, Tianyi and Hosseini, Mohammad Javad and Weber, Sabine and Steedman, Mark},
  year = {2025},
  pages = {903–921},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-110866x},
  url = {https://fis.uni-bamberg.de/handle/uniba/110866},
  language = {eng},
  abstract = {We examine LMs’ competence of directional predicate entailments by supervised fine-tuning with prompts. Our analysis shows that contrary to their apparent success on standard NLI, LMs show limited ability to learn such directional inference; moreover, existing datasets fail to test directionality, and/or are infested by artefacts that can be learnt as proxy for entailments, yielding over-optimistic results. In response, we present BoOQA (Boolean Open QA), a robust multi-lingual evaluation benchmark for directional predicate entailments, extrinsic to existing training sets. On BoOQA, we establish baselines and show evidence of existing LM-prompting models being incompetent directional entailment learners, in contrast to entailment graphs, however limited by sparsity.},
  keywords = {Language Models}
}

@inproceedings{li2025_110858,
  title = {Cross-lingual Inference with A Chinese Entailment Graph},
  author = {Li, Tianyi and Weber, Sabine and Hosseini, Mohammad Javad and Guillou, Liane and Steedman, Mark},
  year = {2025},
  pages = {1214–1233},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-110858x},
  url = {https://fis.uni-bamberg.de/handle/uniba/110858},
  language = {eng},
  abstract = {Predicate entailment detection is a crucial task for question-answering from text, where previous work has explored unsupervised learning of entailment graphs from typed open relation triples. In this paper, we present the first pipeline for building Chinese entailment graphs, which involves a novel high-recall open relation extraction (ORE) method and the first Chinese fine-grained entity typing dataset under the FIGER type ontology. Through experiments on the Levy-Holt dataset, we verify the strength of our Chinese entailment graph, and reveal the cross-lingual complementarity: on the parallel Levy-Holt dataset, an ensemble of Chinese and English entailment graphs outperforms both monolingual graphs, and raises unsupervised SOTA by 4.7 AUC points.},
  keywords = {Chinese Entailment Graph}
}

@inproceedings{weber2025_110857,
  title = {Fine-grained General Entity Typing in German using GermaNet},
  author = {Weber, Sabine and Steedman, Mark},
  year = {2025},
  pages = {138–143},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-110857x},
  url = {https://fis.uni-bamberg.de/handle/uniba/110857},
  language = {eng},
  abstract = {Fine-grained entity typing is important to tasks like relation extraction and knowledge base construction. We find however, that fine-grained entity typing systems perform poorly on general entities (e.g. “ex-president”) as compared to named entities (e.g. “Barack Obama”). This is due to a lack of general entities in existing training data sets. We show that this problem can be mitigated by automatically generating training data from WordNets. We use a German WordNet equivalent, GermaNet, to automatically generate training data for German general entity typing. We use this data to supplement named entity data to train a neural fine-grained entity typing system. This leads to a 10\% improvement in accuracy of the prediction of level 1 FIGER types for German general entities, while decreasing named entity type prediction accuracy by only 1\%.},
  keywords = {GermaNet}
}

@inproceedings{weber2025_110856,
  title = {Zero-Shot Cross-Lingual Transfer is a Hard Baseline to Beat in German Fine-Grained Entity Typing},
  author = {Weber, Sabine and Steedman, Mark},
  year = {2025},
  pages = {42-48},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-110856x},
  url = {https://fis.uni-bamberg.de/handle/uniba/110856},
  language = {eng},
  abstract = {The training of NLP models often requires large amounts of labelled training data, which makes it difficult to expand existing models to new languages. While zero-shot cross-lingual transfer relies on multilingual word embeddings to apply a model trained on one language to another, Yarowski and Ngai (2001) propose the method of annotation projection to generate training data without manual annotation. This method was successfully used for the tasks of named entity recognition and coarse-grained entity typing, but we show that it is outperformed by zero-shot cross-lingual transfer when applied to the similar task of fine-grained entity typing. In our study of fine-grained entity typing with the FIGER type ontology for German, we show that annotation projection amplifies the English model’s tendency to underpredict level 2 labels and is beaten by zero-shot cross-lingual transfer on three novel test sets.},
  keywords = {Zero-Shot Cross-Lingual Transfer}
}

@inproceedings{srivastava2025_110847,
  title = {Different German and English Coreference Resolution Models for Multi-domain Content Curation Scenarios},
  author = {Srivastava, Ankit and Weber, Sabine and Bourgonje, Peter and Rehm, Georg},
  year = {2025},
  pages = {48-61},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-110847x},
  url = {https://fis.uni-bamberg.de/handle/uniba/110847},
  language = {eng},
  abstract = {Coreference Resolution is the process of identifying all words and phrases in a text that refer to the same entity. It has proven to be a useful intermediary step for a number of natural language processing applications. In this paper, we describe three implementations for performing coreference resolution: rule-based, statistical, and projection-based (from English to German). After a comparative evaluation on benchmark datasets, we conclude with an application of these systems on German and English texts from different scenarios in digital curation such as an archive of personal letters, excerpts from a museum exhibition, and regional news articles.},
  keywords = {Multi-domain Content Curation Scenarios}
}

@inproceedings{schafer2025_109541,
  title = {Which Demographics do LLMs Default to During Annotation?},
  author = {Schäfer, Johannes and Combs, Aidan and Bagdon, Christopher Doyle and Li, Jiahui and Probol, Nadine and Greschner, Lynn and Papay, Sean and Menchaca Resendiz, Yarik and Velutharambath, Aswathy and Wuehrl, Amelie and Weber, Sabine and Klinger, Roman},
  year = {2025},
  booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics},
  pages = {17331–17348},
  publisher = {Association for Computational Linguistics},
  isbn = {979-8-89176-251-0},
  doi = {10.18653/v1/2025.acl-long.848},
  url = {https://fis.uni-bamberg.de/handle/uniba/109541},
  language = {eng},
  abstract = {Demographics and cultural background of annotators influence the labels they assign in text annotation – for instance, an elderly woman might find it offensive to read a message addressed to a “bro”, but a male teenager might find it appropriate. It is therefore important to acknowledge label variations to not underrepresent members of a society. Two research directions developed out of this observation in the context of using large language models (LLM) for data annotations, namely (1) studying biases and inherent knowledge of LLMs and (2) injecting diversity in the output by manipulating the prompt with demographic information. We combine these two strands of research and ask the question to which demographics an LLM resorts when no demographics is given. To answer this question, we evaluate which attributes of human annotators LLMs inherently mimic. Furthermore, we compare non-demographic conditioned prompts and placebo-conditioned prompts (e.g., “you are an annotator who lives in house number 5”) to demographics-conditioned prompts (“You are a 45 year old man and an expert on politeness annotation. How do you rate \{instance\}”). We study these questions for politeness and offensiveness annotations on the POPQUORN data set, a corpus created in a controlled manner to investigate human label variations based on demographics which has not been used for LLM-based analyses so far. We observe notable influences related to gender, race, and age in demographic prompting, which contrasts with previous studies that found no such effects},
  keywords = {-}
}

@inproceedings{li2022_97222,
  title = {Language Models Are Poor Learners of Directional Inference},
  author = {Li, Tianyi and Hosseini, Mohammad Javad and Weber, Sabine and Steedman, Mark},
  year = {2022},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2022},
  pages = {903–921},
  publisher = {Association for Computational Linguistics},
  doi = {10.18653/v1/2022.findings-emnlp.64},
  url = {https://fis.uni-bamberg.de/handle/uniba/97222},
  language = {eng},
  abstract = {We examine LMs’ competence of directional predicate entailments by supervised fine-tuning with prompts. Our analysis shows that contrary to their apparent success on standard NLI, LMs show limited ability to learn such directional inference; moreover, existing datasets fail to test directionality, and/or are infested by artefacts that can be learnt as proxy for entailments, yielding over-optimistic results. In response, we present BoOQA (Boolean Open QA), a robust multi-lingual evaluation benchmark for directional predicate entailments, extrinsic to existing training sets. On BoOQA, we establish baselines and show evidence of existing LM-prompting models being incompetent directional entailment learners, in contrast to entailment graphs, however limited by sparsity.},
  keywords = {Language Models}
}

@inproceedings{li2022_97224,
  title = {Cross-lingual Inference with A Chinese Entailment Graph},
  author = {Li, Tianyi and Weber, Sabine and Hosseini, Mohammad Javad and Guillou, Liane and Steedman, Mark},
  year = {2022},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2022},
  pages = {1214–1233},
  publisher = {Association for Computational Linguistics},
  doi = {10.18653/v1/2022.findings-acl.96},
  url = {https://fis.uni-bamberg.de/handle/uniba/97224},
  language = {eng},
  abstract = {Predicate entailment detection is a crucial task for question-answering from text, where previous work has explored unsupervised learning of entailment graphs from typed open relation triples. In this paper, we present the first pipeline for building Chinese entailment graphs, which involves a novel high-recall open relation extraction (ORE) method and the first Chinese fine-grained entity typing dataset under the FIGER type ontology. Through experiments on the Levy-Holt dataset, we verify the strength of our Chinese entailment graph, and reveal the cross-lingual complementarity: on the parallel Levy-Holt dataset, an ensemble of Chinese and English entailment graphs outperforms both monolingual graphs, and raises unsupervised SOTA by 4.7 AUC points.},
  keywords = {Chinese Entailment Graph}
}

@inproceedings{weber2021_97225,
  title = {Fine-grained General Entity Typing in German using GermaNet},
  author = {Weber, Sabine and Steedman, Mark},
  year = {2021},
  booktitle = {Proceedings of the Fifteenth Workshop on Graph-Based Methods for Natural Language Processing (TextGraphs-15)},
  pages = {138–143},
  publisher = {Association for Computational Linguistics},
  doi = {10.18653/v1/2021.textgraphs-1.14},
  url = {https://fis.uni-bamberg.de/handle/uniba/97225},
  language = {eng},
  abstract = {Fine-grained entity typing is important to tasks like relation extraction and knowledge base construction. We find however, that fine-grained entity typing systems perform poorly on general entities (e.g. “ex-president”) as compared to named entities (e.g. “Barack Obama”). This is due to a lack of general entities in existing training data sets. We show that this problem can be mitigated by automatically generating training data from WordNets. We use a German WordNet equivalent, GermaNet, to automatically generate training data for German general entity typing. We use this data to supplement named entity data to train a neural fine-grained entity typing system. This leads to a 10\% improvement in accuracy of the prediction of level 1 FIGER types for German general entities, while decreasing named entity type prediction accuracy by only 1\%.},
  keywords = {GermaNet}
}

@inproceedings{weber2021_97228,
  title = {Zero-Shot Cross-Lingual Transfer is a Hard Baseline to Beat in German Fine-Grained Entity Typing},
  author = {Weber, Sabine and Steedman, Mark},
  year = {2021},
  booktitle = {Proceedings of the Second Workshop on Insights from Negative Results in NLP},
  pages = {42–48},
  publisher = {Association for Computational Linguistics},
  doi = {10.18653/v1/2021.insights-1.7},
  url = {https://fis.uni-bamberg.de/handle/uniba/97228},
  language = {eng},
  abstract = {The training of NLP models often requires large amounts of labelled training data, which makes it difficult to expand existing models to new languages. While zero-shot cross-lingual transfer relies on multilingual word embeddings to apply a model trained on one language to another, Yarowski and Ngai (2001) propose the method of annotation projection to generate training data without manual annotation. This method was successfully used for the tasks of named entity recognition and coarse-grained entity typing, but we show that it is outperformed by zero-shot cross-lingual transfer when applied to the similar task of fine-grained entity typing. In our study of fine-grained entity typing with the FIGER type ontology for German, we show that annotation projection amplifies the English model’s tendency to underpredict level 2 labels and is beaten by zero-shot cross-lingual transfer on three novel test sets.},
  keywords = {Zero-Shot Cross-Lingual Transfer}
}

@inproceedings{weber2019_97229,
  title = {Construction and Alignment of Multilingual Entailment Graphs for Semantic Inference},
  author = {Weber, Sabine and Steedman, Mark},
  year = {2019},
  booktitle = {Proceedings of the 2019 Workshop on Widening NLP},
  pages = {77–79},
  publisher = {Association for Computational Linguistics},
  url = {https://fis.uni-bamberg.de/handle/uniba/97229},
  language = {eng}
}

@inproceedings{fahndrich2018_97230,
  title = {A Marker Passing Approach to Winograd Schemas},
  author = {Fähndrich, Johannes and Weber, Sabine and Kanthak, Hannes},
  year = {2018},
  booktitle = {Semantic technology : 8th joint international conference, JIST 2018, Awaji, Japan, November 26–28, 2018, proceedings},
  pages = {165-181},
  publisher = {Springer International Publishing},
  address = {Cham, Switzerland},
  isbn = {9783030042837},
  doi = {10.1007/978-3-030-04284-4\_12},
  url = {https://fis.uni-bamberg.de/handle/uniba/97230},
  language = {eng}
}

@inproceedings{srivastava2018_97231,
  title = {Different German and English Coreference Resolution Models for Multi-domain Content Curation Scenarios},
  author = {Srivastava, Ankit and Weber, Sabine and Bourgonje, Peter and Rehm, Georg},
  year = {2018},
  booktitle = {Language Technologies for the Challenges of the Digital Age : 27th International Conference, GSCL 2017, Berlin, Germany, September 13-14, 2017, Proceedings},
  pages = {48-61},
  publisher = {Springer International Publishing},
  address = {Cham},
  isbn = {9783319737058},
  doi = {10.1007/978-3-319-73706-5\_5},
  url = {https://fis.uni-bamberg.de/handle/uniba/97231},
  language = {eng},
  abstract = {Coreference Resolution is the process of identifying all words and phrases in a text that refer to the same entity. It has proven to be a useful intermediary step for a number of natural language processing applications. In this paper, we describe three implementations for performing coreference resolution: rule-based, statistical, and projection-based (from English to German). After a comparative evaluation on benchmark datasets, we conclude with an application of these systems on German and English texts from different scenarios in digital curation such as an archive of personal letters, excerpts from a museum exhibition, and regional news articles.},
  keywords = {Multi-domain Content Curation Scenarios}
}

