Papers by Sebastian Padó
DEbateNet-mig15:Tracing the 2015 Immigration Debate in Germany Over Time (2020.lrec-1)
Copied to clipboard
Gabriella Lapesa, Andre Blessing, Nico Blokker, Erenay Dayanik, Sebastian Haunss, Jonas Kuhn, Sebastian Padó
| Challenge: | a dataset for germany covering the public debate on immigration is annotated . a political science notion of a claim is used to represent the political discourse . |
| Approach: | They annotate a dataset for german public debate on immigration in 2015 using a political science notion of a claim . they identify claims in newspaper articles, assign them to actors and fine-grained categories and annotize their polarity and date. |
| Outcome: | The dataset is annotated by a political science framework and shows it captures political debate . it shows that political actors can change their positions and take a strong stand against them . |
Lexical Substitution for Evaluating Compositional Distributional Models (N18-2)
Copied to clipboard
| Challenge: | Compositional Distributional Semantic Models (CDSMs) model the meaning of phrases and sentences in vector space. |
| Approach: | They propose to use lexical substitution to evaluate CDSMs by comparing a LexSub-annotated corpus with a manual LexSub annotation. |
| Outcome: | The proposed model outperforms simple component-wise CDSMs and performs on par with the context2vec LexSub model using the same context. |
Similar, but why? A Toolkit for Explaining Text Similarity (2026.eacl-demo)
Copied to clipboard
| Challenge: | XPLAINSIM is a Python package that explains textual similarity in an easy-to-use way. |
| Approach: | They propose a Python package that unifies three approaches to explain text similarity . they demonstrate the value of the package through intuitive examples and empirical research . |
| Outcome: | XPLAINSIM is a Python package that unifies three approaches to explain text similarity . the authors show that the package is useful for explaining text similarities in a simple way . |
One Persona, Many Cues, Different Results: How Sociodemographic Cues Impact LLM Personalization (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work has used personas to study biases by relying on a single cue to prompt a persona, such as user names or explicit attribute mentions. |
| Approach: | They compare six commonly used personacues across seven open and proprietary LLMs on four writing and advice tasks. |
| Outcome: | The proposed model is based on a persona, a synthetic user profile defined by specific attributes, defined by gender or race. |
Masking Actor Information Leads to Fairer Political Claims Detection (2020.acl-main)
Copied to clipboard
| Challenge: | In recent years, NLP methods have found increasing adoption in the social sciences . however, CSS must be crucially interested in the algorithmic fairness of the underlying methods . |
| Approach: | They propose two methods which mask proper names and pronouns during training of the model, thus removing personal information bias. |
| Outcome: | The proposed methods decrease frequency bias while keeping the overall performance stable. |
DERE: A Task and Domain-Independent Slot Filling Framework for Declarative Relation Extraction (D18-2)
Copied to clipboard
| Challenge: | Comparability of models across tasks is lacking in most machine learning systems for natural language processing. |
| Approach: | They propose a framework for declarative specification and compilation of template-based information extraction that uses a generic specification language for the task and for data annotations in terms of spans and frames. |
| Outcome: | The proposed framework enables representation of a large variety of natural language processing tasks. |
Towards Understanding the Relationship between In-context Learning and Compositional Generalization (2024.lrec-main)
Copied to clipboard
| Challenge: | In-context learning is an inductive bias for compositional generalization, but many deep neural architectures struggle with this ability. |
| Approach: | They propose to force a causal Transformer to in-context learn to promote compositional generalization by using earlier examples to generalize to later ones. |
| Outcome: | The proposed model can solve 'ordinary' learning problems by utilizing earlier examples to generalize to later ones, i.e., in-context learning. |
A Named Entity Recognition Shootout for German (P18-2)
Copied to clipboard
| Challenge: | Named entity recognition and classification (NER) is a central component in many natural language processing pipelines. |
| Approach: | They propose to build a model for German named entity recognition that performs at the state of the art for both contemporary and historical texts. |
| Outcome: | The proposed model outperforms the CRF and BiLSTM on large and small datasets. |
RiQuA: A Corpus of Rich Quotation Annotation for English Literary Text (2020.lrec-1)
Copied to clipboard
| Challenge: | In literature, spoken interactions between characters are of central importance to the narrative. |
| Approach: | They propose to annotate quotations, including their interpersonal structure, for English literary text. |
| Outcome: | The proposed dataset provides a rich view of dialogue structures not available from other available corpora. |
Lost in Back-Translation: Emotion Preservation in Neural Machine Translation (2020.coling-main)
Copied to clipboard
| Challenge: | MT is used to support human-to-human communication across languages, but it is unclear whether it can translate the non-propositional level of emotions. |
| Approach: | They propose to use a re-ranking approach to change emotions to reverse this tendency . they find that emotions are toned down or amplified through linguistic changes . |
| Outcome: | The proposed model can be used to change emotions, and it can be applied to other languages. |
Multi-Dimensional Machine Translation Evaluation: Model Evaluation and Resource for Korean (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing studies on MT evaluation characterize quality of output with a single number . a recent advancement in MT technologies has enabled higher-quality, more nuanced translations . |
| Approach: | They propose a 1200-sentence MQM evaluation benchmark for English-Korean and a reference-free QE setup to evaluate the quality of the translations. |
| Outcome: | The proposed model outperforms the existing model in style and accuracy. |
Interpretable Text Embeddings and Text Similarity Explanation: A Survey (2025.emnlp-main)
Copied to clipboard
| Challenge: | Text embeddings are a fundamental component in many NLP tasks, but their interpretation and explanation remain challenging. |
| Approach: | They propose a framework for interpretable text embeddings and text similarity explanation . they characterize the main ideas, approaches, and trade-offs and discuss lessons learned . |
| Outcome: | The proposed methods are compared with existing models and compare them with existing ones. |
Media Bias Detection Across Families of Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Traditional NLP models have shown good performance in classifying media bias, but require careful model design and extensive tuning. |
| Approach: | They ask how well prompting of large language models can recognize media bias. |
| Outcome: | The prompt-based models deliver comparable performance to traditional models with greatly reduced effort and the availability of context substantially improves results. |
Representation biases in sentence transformers (2023.eacl-main)
Copied to clipboard
| Challenge: | argued that transformer-based models are not well suited for sentence-level downstream tasks. |
| Approach: | They propose to use sentence transformers to produce full-sentence representations . they propose to combine transformers with a training regime that embeds tokens into the model . |
| Outcome: | The proposed model performs better on downstream tasks than the vanilla model and its variants. |
An Attribution Method for Siamese Encoders (2023.emnlp-main)
Copied to clipboard
| Challenge: | Despite the success of Siamese encoder models, little is known about the aspects of inputs they base their decisions on. |
| Approach: | They propose a local attribution method for Siamese encoders by generalizing the principle of integrated gradients to models with multiple inputs. |
| Outcome: | The proposed method can be reduced to a token–token matrix and account for the model’s full computation graph and is guaranteed to converge to the actual prediction. |
Approximate Attributions for Off-the-Shelf Siamese Transformers (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing attribution methods cannot tackle Siamese encoders since they compare two inputs rather than processing a single one. |
| Approach: | They propose an attribution method specifically targeted for Siamese encoders that can be adjusted and fine-tuned to retain original model's predictive performance. |
| Outcome: | The proposed method retains the original model's predictive performance and can be applied to off-the-shelf models. |
Who Sides with Whom? Towards Computational Construction of Discourse Networks for Political Debates (P19-1)
Copied to clipboard
| Challenge: | a vision of computational construction of discourse networks from newspaper reports is essential for understanding democratic political decision making. |
| Approach: | They propose to use a requirements analysis and an annotated pilot corpus of migration claims to build a computationally-based model of political debates from newspaper reports. |
| Outcome: | The proposed framework could be scaled up to a large scale and be useful for political scientists. |
Adverbs, Surprisingly (2023.starsem-1)
Copied to clipboard
| Challenge: | adverbs are the part of speech (POS) that has seen the least attention in computational linguistics due to its challenging nature. |
| Approach: | They propose to use Frame Semantics to characterize word meaning to uncover systematic gaps in adverb accounts. |
| Outcome: | The proposed approach can describe ambiguity, semantic roles, and null instantiation of adverbs. |
Multilingual estimation of political-party positioning: From label aggregation to long-input Transformers (2023.emnlp-main)
Copied to clipboard
| Challenge: | Scaling analysis is a technique that assigns a political actor a score on a predefined scale based on 'typically long' text. |
| Approach: | They propose to use label aggregation and long-input-Transformer-based models to automatically scale political-party manifestos. |
| Outcome: | The proposed models can scale political platforms on a predefined scale based on 'left-right' scales and work robustly across domains and languages. |
An Environment for Relational Annotation of Political Debates (P19-3)
Copied to clipboard
| Challenge: | Scalable text analysis techniques can open corpora to new questions in computational social sciences and digital humanities. |
| Approach: | They describe a tool that allows annotating newspaper text with rich information about claims (demands) raised by politicians and other actors. |
| Outcome: | The MARDY tool realizes the complete workflow necessary for annotating a large newspaper text collection with rich information about claims (demands) raised by politicians and other actors. |
Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing work on marginal distributions and model steering fails to account for deeper latent structures that characterise real populations. |
| Approach: | They propose a framework for evaluating the representativeness of aligned models through multivariate correlation patterns in addition to marginal distributions. |
| Outcome: | The proposed framework compares two model steering techniques against human responses from the World Values Survey. |
Do Political Opinions Transfer Between Western Languages? An Analysis of Unaligned and Aligned Multilingual LLMs (2026.eacl-long)
Copied to clipboard
| Challenge: | Public opinion surveys show cross-cultural differences in political opinions between socio-cultural contexts. |
| Approach: | They analyze whether opinions transfer between languages or whether there are separate opinions for each language in multilingual large language models of various sizes across five Western languages. |
| Outcome: | The political alignment shifts opinions almost uniformly across all five languages. |
Toeing the Party Line: Election Manifestos as a Key to Understand Political Discourse on Twitter (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work on political positioning on Twitter has tended to focus on manifestos rather than social media since it is ambiguous and dependent on social context. |
| Approach: | They propose to use hashtags as a signal to fine-tune text representations for politicians' tweets using a hashtag-based method to predict pairwise positional similarities between parties from the manifesto case to the Twitter case. |
| Outcome: | The proposed method matches politicians' statements to official lines of the parties' tweets, even when only small subsets from shorter time periods are available. |
Additive manifesto decomposition: A policy domain aware method for understanding party positioning (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for estimating policy domain aware party similarity are limited to global domains. |
| Approach: | They propose a workflow for estimating policy domain aware party similarity by aggregating policy domains into a single figure . they use a set of tools to extract party positions on major policy axes via multidimensional scaling. |
| Outcome: | The proposed method yields high correlation when predicting party similarity at a global level and provides accurate party-specific positions even with automatically labelled policy domains. |
Crowdsourcing and Validating Event-focused Emotion Corpora for German and English (P19-1)
Copied to clipboard
| Challenge: | Existing studies on automatic recognition of emotions in text have achieved promising results, but there is a shortage of resources for non-English languages, with few exceptions, like Chinese. |
| Approach: | They propose to use a crowdsourced German emotion corpus to build a corpus similar to the English ISEAR emotion dataset. |
| Outcome: | The proposed model performs well in German and English, but lacks the resources for non-English languages. |
Dissecting Span Identification Tasks with Performance Prediction (2020.emnlp-main)
Copied to clipboard
| Challenge: | Span identification tasks are a staple of applied NLP, but there is little insight on how their properties influence their difficulty. |
| Approach: | They propose to build a model to predict span ID performance for unseen span ID tasks that can support architecture choices. |
| Outcome: | The proposed model predicts span ID tasks for unseen span ID task in English, and the meta model predictable span ID performance. |