Sebastian Padó

Dmitry Nikolaev, Sebastian Padó

Variants of the BERT architecture specialised for producing full-sentence representations often achieve better performance on downstream tasks than sentence embeddings extracted from vanilla BERT. However, there is still little understanding of what properties of inputs determine the properties of such representations. In this study, we construct several sets of sentences with pre-defined lexical and syntactic structures and show that SOTA sentence transformers have a strong nominal-participant-set bias: cosine similarities between pairs of sentences are more strongly determined by the overlap in the set of their noun participants than by having the same predicates, lengthy nominal modifiers, or adjuncts. At the same time, the precise syntactic-thematic functions of the participants are largely irrelevant.

Tanise Ceron, Nico Blokker, Sebastian Padó

Even though fine-tuned neural language models have been pivotal in enabling "deep" automatic text analysis, optimizing text representations for specific applications remains a crucial bottleneck. In this study, we look at this problem in the context of a task from computational social science, namely modeling pairwise similarities between political parties. Our research question is what level of structural information is necessary to create robust text representation, contrasting a strongly informed approach (which uses both claim span and claim category annotations) with approaches that forgo one or both types of annotation with document structure-based heuristics. Evaluating our models on the manifestos of German parties for the 2021 federal election. We find that heuristics that maximize within-party over between-party similarity along with a normalization step lead to reliable party similarity prediction, without the need for manual annotation.

Dmitry Nikolaev, Sebastian Padó

The capabilities and limitations of BERT and similar models are still unclear when it comes to learning syntactic abstractions, in particular across languages. In this paper, we use the task of subordinate-clause detection within and across languages to probe these properties. We show that this task is deceptively simple, with easy gains offset by a long tail of harder cases, and that BERT's zero-shot performance is dominated by word-order effects, mirroring the SVO/VSO/SOV typology.

Lucas Möller, Pascal Tilli, Ngoc Thang Vu et al.

Dual encoder architectures like Clip models map two types of inputs into a shared embedding space and predict similarities between them. Despite their wide application, it is, however, not understood how these models compare their two inputs. Common first-order feature-attribution methods explain importances of individual features and can, thus, only provide limited insights into dual encoders, whose predictions depend on interactions between features. In this paper, we first derive a second-order method enabling the attribution of predictions by any differentiable dual encoder onto feature-interactions between its inputs. Second, we apply our method to Clip models and show that they learn fine-grained correspondences between parts of captions and regions in images. They match objects across input modes and also account for mismatches. This intrinsic visual-linguistic grounding ability, however, varies heavily between object classes, exhibits pronounced out-of-domain effects and we can identify individual errors as well as systematic failure categories. Code is publicly available: https://github.com/lucasmllr/exCLIP

Lucas Möller, Sebastian Padó

A number of models for neural content-based news recommendation have been proposed. However, there is limited understanding of the relative importances of the three main components of such systems (news encoder, user encoder, and scoring function) and the trade-offs involved. In this paper, we assess the hypothesis that the most widely used means of matching user and candidate news representations is not expressive enough. We allow our system to model more complex relations between the two by assessing more expressive scoring functions. Across a wide range of baseline and established systems this results in consistent improvements of around 6 points in AUC. Our results also indicate a trade-off between the complexity of news encoder and scoring function: A fairly simple baseline model scores well above 68% AUC on the MIND dataset and comes within 2 points of the published state-of-the-art, while requiring a fraction of the computational costs.

Hardy, Sebastian Padó, Amelie Wührl et al.

News Recommender Systems (NRS) shape what users read, whose perspectives they encounter, and influence public discourse. Yet their design is value-laden: intentionally or not, NRS can embed undesired values in recommendation procedures, such as excluding underrepresented voices or favoring certain viewpoints, which may conflict with democratic goals. Existing solutions also lack mechanisms to explicitly control these values. Therefore, we introduce an approach that parameterizes NRS to support different democratic goals. We propose Aspect-Aware Candidate Generation (A2CG), a normatively configurable procedure for the candidate generation stage of NRS that allows designers to shape diversity in recommendations. Unlike prior work that only re-ranks candidates, A2CG introduces diversity at the start of the recommendation pipeline. A2CG represents articles along multiple diversity aspects: sentiment, political leaning, topic, and media framing. User interests are encoded using a Vector Quantized VAE, while a decoder-only model predicts the next article aspects users are likely to engage with. To broaden exposure to perspectives, A2CG injects diversity during retrieval by selectively flipping aspects in the predicted query, allowing candidate diversity to be tuned toward specific democratic models. Our method enables normative configurations that existing NRS cannot express. Unlike baselines with fixed structural biases, A2CG supports continuous calibration between democratic ideals without retraining. Empirically, A2CG generates novel, diverse, and serendipitous candidates while providing explicit parameter-driven control over the trade-off between personalization and democratic alignment. Rather than aiming for pointwise superiority, A2CG's main contribution lies in its controllability and ability to express flexible normative configurations.

Dmitry Nikolaev, Sebastian Padó

The question of what kinds of linguistic information are encoded in different layers of Transformer-based language models is of considerable interest for the NLP community. Existing work, however, has overwhelmingly focused on word-level representations and encoder-only language models with the masked-token training objective. In this paper, we present experiments with semantic structural probing, a method for studying sentence-level representations via finding a subspace of the embedding space that provides suitable task-specific pairwise distances between data-points. We apply our method to language models from different families (encoder-only, decoder-only, encoder-decoder) and of different sizes in the context of two tasks, semantic textual similarity and natural-language inference. We find that model families differ substantially in their performance and layer dynamics, but that the results are largely model-size invariant.

Lucas Möller, Dmitry Nikolaev, Sebastian Padó

Despite the success of Siamese encoder models such as sentence transformers (ST), little is known about the aspects of inputs they pay attention to. A barrier is that their predictions cannot be attributed to individual features, as they compare two inputs rather than processing a single one. This paper derives a local attribution method for Siamese encoders by generalizing the principle of integrated gradients to models with multiple inputs. The solution takes the form of feature-pair attributions, and can be reduced to a token-token matrix for STs. Our method involves the introduction of integrated Jacobians and inherits the advantageous formal properties of integrated gradients: it accounts for the model's full computation graph and is guaranteed to converge to the actual prediction. A pilot study shows that in an ST few token-pairs can often explain large fractions of predictions, and it focuses on nouns and verbs. For accurate predictions, it however needs to attend to the majority of tokens and parts of speech.

Dmitry Nikolaev, Tanise Ceron, Sebastian Padó

Scaling analysis is a technique in computational political science that assigns a political actor (e.g. politician or party) a score on a predefined scale based on a (typically long) body of text (e.g. a parliamentary speech or an election manifesto). For example, political scientists have often used the left--right scale to systematically analyse political landscapes of different countries. NLP methods for automatic scaling analysis can find broad application provided they (i) are able to deal with long texts and (ii) work robustly across domains and languages. In this work, we implement and compare two approaches to automatic scaling analysis of political-party manifestos: label aggregation, a pipeline strategy relying on annotations of individual statements from the manifestos, and long-input-Transformer-based models, which compute scaling values directly from raw text. We carry out the analysis of the Comparative Manifestos Project dataset across 41 countries and 27 languages and find that the task can be efficiently solved by state-of-the-art models, with label aggregation producing the best results.

Urs Zaberer, Sebastian Padó, Gabriella Lapesa

The identification and classification of political claims is an important step in the analysis of political newspaper reports; however, resources for this task are few and far between. This paper explores different strategies for the cross-lingual projection of political claims analysis. We conduct experiments on a German dataset, DebateNet2.0, covering the policy debate sparked by the 2015 refugee crisis. Our evaluation involves two tasks (claim identification and categorization), three languages (German, English, and French) and two methods (machine translation -- the best method in our experiments -- and multilingual embeddings).

Hardy, Sebastian Padó

Large Language Models (LLMs) achieve strong linguistic performance, yet their internal mechanisms for producing these predictions remain unclear. We investigate the hypothesis that LLMs encode representations of linguistic constraint violations within their parameters, which are selectively activated when processing ungrammatical sentences. To test this, we use sparse autoencoders to decompose polysemantic activations into sparse, monosemantic features and recover candidates for violation-related features. We introduce a sensitivity score for identifying features that are preferentially activated on constraint-violated versus well-formed inputs, enabling unsupervised detection of potential violation-specific features. We further propose a conjunctive falsification framework with three criteria evaluated jointly. Overall, the results are negative in two respects: (1) the falsification criteria are not jointly satisfied across linguistic phenomena, and (2) no features are consistently shared across all categories. While some phenomena show partial evidence of selective causal structure, the overall pattern provides limited support for a unified set of grammatical violation detectors in current LMs.

Tristan Williams, Franziska Weeber, Sebastian Padó et al.

Large language models are increasingly used to represent human opinions, values, or beliefs, and their steerability towards these ideals is an active area of research. Existing work focuses predominantly on aligning marginal response distributions, treating each survey item independently. While essential, this may overlook deeper latent structures that characterise real populations and underpin cultural values theories. We propose a framework for evaluating the representativeness of aligned models through multivariate correlation patterns in addition to marginal distributions. We show the value of our evaluation scheme by comparing two model steering techniques (persona prompting and demographic fine-tuning) and evaluating them against human responses from the World Values Survey. While the demographically fine-tuned model better approximates marginal response distributions than persona prompting, both techniques fail to fully capture the gold standard correlation patterns. We conclude that representativeness is a distinct aspect of value alignment and an evaluation focused on marginals can mask structural failures, leading to overly optimistic conclusions about model capabilities.

Franziska Weeber, Vera Neplenbroek, Jan Batzner et al.

Personalization of LLMs by sociodemographic subgroup often improves user experience, but can also introduce or amplify biases and unfair outcomes across groups. Prior work has employed so-called personas, sociodemographic user attributes conveyed to a model, to study bias in LLMs by relying on a single cue to prompt a persona, such as user names or explicit attribute mentions. This disregards LLM sensitivity to prompt variations (robustness) and the rarity of some cues in real interactions (external validity). We compare six commonly used persona cues across seven open and proprietary LLMs on four writing and advice tasks. While cues are overall highly correlated, they produce substantial variance in responses across personas. We therefore caution against claims from a single persona cue and recommend future personalization research to evaluate multiple externally valid cues.

Tanise Ceron, Neele Falk, Ana Barić et al.

Due to the widespread use of large language models (LLMs), we need to understand whether they embed a specific "worldview" and what these views reflect. Recent studies report that, prompted with political questionnaires, LLMs show left-liberal leanings (Feng et al., 2023; Motoki et al., 2024). However, it is as yet unclear whether these leanings are reliable (robust to prompt variations) and whether the leaning is consistent across policies and political leaning. We propose a series of tests which assess the reliability and consistency of LLMs' stances on political statements based on a dataset of voting-advice questionnaires collected from seven EU countries and annotated for policy issues. We study LLMs ranging in size from 7B to 70B parameters and find that their reliability increases with parameter count. Larger models show overall stronger alignment with left-leaning parties but differ among policy programs: They show a (left-wing) positive stance towards environment protection, social welfare state and liberal society but also (right-wing) law and order, with no consistent preferences in the areas of foreign policy and migration.

Lucas Möller, Dmitry Nikolaev, Sebastian Padó

Siamese encoders such as sentence transformers are among the least understood deep models. Established attribution methods cannot tackle this model class since it compares two inputs rather than processing a single one. To address this gap, we have recently proposed an attribution method specifically for Siamese encoders (Möller et al., 2023). However, it requires models to be adjusted and fine-tuned and therefore cannot be directly applied to off-the-shelf models. In this work, we reassess these restrictions and propose (i) a model with exact attribution ability that retains the original model's predictive performance and (ii) a way to compute approximate attributions for off-the-shelf models. We extensively compare approximate and exact attributions and use them to analyze the models' attendance to different linguistic aspects. We gain insights into which syntactic roles Siamese transformers attend to, confirm that they mostly ignore negation, explore how they judge semantically opposite adjectives, and find that they exhibit lexical bias.

Katrina Olsen, Sebastian Padó

Nonsensical and anomalous sentences have been instrumental in the development of computational models of semantic interpretation. A core challenge is to distinguish between what is merely anomalous (but can be interpreted given a supporting context) and what is truly nonsensical. However, it is unclear (a) how nonsensical, rather than merely anomalous, existing datasets are; and (b) how well LLMs can make this distinction. In this paper, we answer both questions by collecting sensicality judgments from human raters and LLMs on sentences from five semantically deviant datasets: both context-free and when providing a context. We find that raters consider most sentences at most anomalous, and only a few as properly nonsensical. We also show that LLMs are substantially skilled in generating plausible contexts for anomalous cases.

Sungjun Han, Sebastian Padó

According to the principle of compositional generalization, the meaning of a complex expression can be understood as a function of the meaning of its parts and of how they are combined. This principle is crucial for human language processing and also, arguably, for NLP models in the face of out-of-distribution data. However, many neural network models, including Transformers, have been shown to struggle with compositional generalization. In this paper, we hypothesize that forcing models to in-context learn can provide an inductive bias to promote compositional generalization. To test this hypothesis, we train a causal Transformer in a setting that renders ordinary learning very difficult: we present it with different orderings of the training instance and shuffle instance labels. This corresponds to training the model on all possible few-shot learning problems attainable from the dataset. The model can solve the task, however, by utilizing earlier examples to generalize to later ones (i.e. in-context learning). In evaluations on the datasets, SCAN, COGS, and GeoQuery, models trained in this manner indeed show improved compositional generalization. This indicates the usefulness of in-context learning problems as an inductive bias for generalization.

Agnese Daffara, Sourabh Dattawad, Sebastian Padó et al.

Frames capture aspects of an issue that are emphasized in a debate by interlocutors and can help us understand how political language conveys different perspectives and ultimately shapes people's opinions. The Media Frame Corpus (MFC) is the most commonly used framework with categories and detailed guidelines for operationalizing frames. It is, however, focused on a few salient U.S. news issues, making it unclear how well these frames can capture news issues in other cultural contexts. To explore this, we introduce FrameNews-PT, a dataset of Brazilian Portuguese news articles covering political and economic news and annotate it within the MFC framework. Through several annotation rounds, we evaluate the extent to which MFC frames generalize to the Brazilian debate issues. We further evaluate how fine-tuned and zero-shot models perform on out-of-domain data. Results show that the 15 MFC frames remain broadly applicable with minor revisions of the guidelines. However, some MFC frames are rarely used, and novel news issues are analyzed using general 'fall-back' frames. We conclude that cross-cultural frame use requires careful consideration.

Ana Barić, Sean Papay, Sebastian Padó

The identification of political actors who put forward claims in public debate is a crucial step in the construction of discourse networks, which are helpful to analyze societal debates. Actor identification is, however, rather challenging: Often, the locally mentioned speaker of a claim is only a pronoun ("He proposed that [claim]"), so recovering the canonical actor name requires discourse understanding. We compare a traditional pipeline of dedicated NLP components (similar to those applied to the related task of coreference) with a LLM, which appears a good match for this generation task. Evaluating on a corpus of German actors in newspaper reports, we find surprisingly that the LLM performs worse. Further analysis reveals that the LLM is very good at identifying the right reference, but struggles to generate the correct canonical form. This points to an underlying issue in LLMs with controlling generated output. Indeed, a hybrid model combining the LLM with a classifier to normalize its output substantially outperforms both initial models.

Juri Opitz, Lucas Möller, Andrianos Michail et al.

Text embeddings are a fundamental component in many NLP tasks, including classification, regression, clustering, and semantic search. However, despite their ubiquitous application, challenges persist in interpreting embeddings and explaining similarities between them. In this work, we provide a structured overview of methods specializing in inherently interpretable text embeddings and text similarity explanation, an underexplored research area. We characterize the main ideas, approaches, and trade-offs. We compare means of evaluation, discuss overarching lessons learned and finally identify opportunities and open challenges for future research.

Franziska Weeber, Tanise Ceron, Sebastian Padó

Public opinion surveys show cross-cultural differences in political opinions between socio-cultural contexts. However, there is no clear evidence whether these differences translate to cross-lingual differences in multilingual large language models (MLLMs). We analyze whether opinions transfer between languages or whether there are separate opinions for each language in MLLMs of various sizes across five Western languages. We evaluate MLLMs' opinions by prompting them to report their (dis)agreement with political statements from voting advice applications. To better understand the interaction between languages in the models, we evaluate them both before and after aligning them with more left or right views using direct preference optimization and English alignment data only. Our findings reveal that unaligned models show only very few significant cross-lingual differences in the political opinions they reflect. The political alignment shifts opinions almost uniformly across all five languages. We conclude that in Western language contexts, political opinions transfer between languages, demonstrating the challenges in achieving explicit socio-linguistic, cultural, and political alignment of MLLMs.

Sebastian Padó, Kerstin Thomas

Emotions are a fundamental aspect of artistic expression. Due to their abstract nature, there is a broad spectrum of emotion realization in artworks. These are subject to historical change and their analysis requires expertise in art history. In this article, we investigate which aspects of emotional expression can be detected by current (2025) vision language models (VLMs). We present a case study of three VLMs (Llava-Llama and two Qwen models) in which we ask these models four sets of questions of increasing complexity about artworks (general content, emotional content, expression of emotions, and emotion symbols) and carry out a qualitative expert evaluation. We find that the VLMs recognize the content of the images surprisingly well and often also which emotions they depict and how they are expressed. The models perform best for concrete images but fail for highly abstract or highly symbolic images. Reliable recognition of symbols remains fundamentally difficult. Furthermore, the models continue to exhibit the well-known LLM weakness of providing inconsistent answers to related questions.

Gabriel Lindenmaier, Sean Papay, Sebastian Padó

Transformer-based language models have recently been at the forefront of active research in text generation. However, these models' advances come at the price of prohibitive training costs, with parameter counts in the billions and compute requirements measured in petaflop/s-decades. In this paper, we investigate transformer-based architectures for improving model performance in a low-data regime by selectively replacing attention layers with feed-forward and quasi-recurrent neural network layers. We test these architectures on the standard Enwik8 and Wikitext-103 corpora. Our results show that our reduced architectures outperform existing models with a comparable number of parameters, and obtain comparable performance to larger models while significantly reducing the number of parameters.

Allison Keith, Antonio Rojas Castro, Hanno Ehrlicher et al.

In theatre, playwrights use the portrayal of characters to explore culturally based gender norms. In this paper, we develop quantitative methods to study gender depiction in the non-religious works (comedias) of Pedro Calderón de la Barca, a prolific Spanish 17th century author. We gather insights from a corpus of more than 100 plays by using a gender classifier and applying model explainability (attribution) methods to determine which text features are most influential in the model's decision to classify speech as 'male' or 'female', indicating the most gendered elements of dialogue in Calderón's comedias in a human accessible manner. We find that female and male characters are portrayed differently and can be identified by the gender prediction model at practically useful accuracies (up to f=0.83). Analysis reveals semantic aspects of gender portrayal, and demonstrates that the model is even useful in providing a relatively accurate scene-by-scene prediction of cross-dressing characters.

Maximilian Maurer, Tanise Ceron, Sebastian Padó et al.

Political discourse on Twitter is a moving target: politicians continuously make statements about their positions. It is therefore crucial to track their discourse on social media to understand their ideological positions and goals. However, Twitter data is also challenging to work with since it is ambiguous and often dependent on social context, and consequently, recent work on political positioning has tended to focus strongly on manifestos (parties' electoral programs) rather than social media. In this paper, we extend recently proposed methods to predict pairwise positional similarities between parties from the manifesto case to the Twitter case, using hashtags as a signal to fine-tune text representations, without the need for manual annotation. We verify the efficacy of fine-tuning and conduct a series of experiments that assess the robustness of our method for low-resource scenarios. We find that our method yields stable positioning reflective of manifesto positioning, both in scenarios with all tweets of candidates across years available and when only smaller subsets from shorter time periods are available. This indicates that it is possible to reliably analyze the relative positioning of actors forgoing manual annotation, even in the noisier context of social media.

Dojun Park, Sebastian Padó

Almost all frameworks for the manual or automatic evaluation of machine translation characterize the quality of an MT output with a single number. An exception is the Multidimensional Quality Metrics (MQM) framework which offers a fine-grained ontology of quality dimensions for scoring (such as style, fluency, accuracy, and terminology). Previous studies have demonstrated the feasibility of MQM annotation but there are, to our knowledge, no computational models that predict MQM scores for novel texts, due to a lack of resources. In this paper, we address these shortcomings by (a) providing a 1200-sentence MQM evaluation benchmark for the language pair English-Korean and (b) reframing MT evaluation as the multi-task problem of simultaneously predicting several MQM scores using SOTA language models, both in a reference-based MT evaluation setup and a reference-free quality estimation (QE) setup. We find that reference-free setup outperforms its counterpart in the style dimension while reference-based models retain an edge regarding accuracy. Overall, RemBERT emerges as the most promising model. Through our evaluation, we offer an insight into the translation quality in a more fine-grained, interpretable manner.

Dmitry Nikolaev, Collin F. Baker, Miriam R. L. Petruck et al.

This paper begins with the premise that adverbs are neglected in computational linguistics. This view derives from two analyses: a literature review and a novel adverb dataset to probe a state-of-the-art language model, thereby uncovering systematic gaps in accounts for adverb meaning. We suggest that using Frame Semantics for characterizing word meaning, as in FrameNet, provides a promising approach to adverb analysis, given its ability to describe ambiguity, semantic roles, and null instantiation.

Tanise Ceron, Dmitry Nikolaev, Sebastian Padó

Automatic extraction of party (dis)similarities from texts such as party election manifestos or parliamentary speeches plays an increasing role in computational political science. However, existing approaches are fundamentally limited to targeting only global party (dis)-similarity: they condense the relationship between a pair of parties into a single figure, their similarity. In aggregating over all policy domains (e.g., health or foreign policy), they do not provide any qualitative insights into which domains parties agree or disagree on. This paper proposes a workflow for estimating policy domain aware party similarity that overcomes this limitation. The workflow covers (a) definition of suitable policy domains; (b) automatic labeling of domains, if no manual labels are available; (c) computation of domain-level similarities and aggregation at a global level; (d) extraction of interpretable party positions on major policy axes via multidimensional scaling. We evaluate our workflow on manifestos from the German federal elections. We find that our method (a) yields high correlation when predicting party similarity at a global level and (b) provides accurate party-specific positions, even with automatically labelled policy domains.

Moiz Rauf, Sebastian Padó, Michael Pradel

Source code summarization is the task of generating a high-level natural language description for a segment of programming language code. Current neural models for the task differ in their architecture and the aspects of code they consider. In this paper, we show that three SOTA models for code summarization work well on largely disjoint subsets of a large code-base. This complementarity motivates model combination: We propose three meta-models that select the best candidate summary for a given code segment. The two neural models improve significantly over the performance of the best individual model, obtaining an improvement of 2.1 BLEU points on a dataset of code segments where at least one of the individual models obtains a non-zero BLEU.

Nico Blokker, André Blessing, Erenay Dayanik et al.

Newspaper reports provide a rich source of information on the unfolding of public debate on specific policy fields that can serve as basis for inquiry in political science. Such debates are often triggered by critical events, which attract public attention and incite the reactions of political actors: crisis sparks the debate. However, due to the challenges of reliable annotation and modeling, few large-scale datasets with high-quality annotation are available. This paper introduces DebateNet2.0, which traces the political discourse on the European refugee crisis in the German quality newspaper taz during the year 2015. The core units of our annotation are political claims (requests for specific actions to be taken within the policy field) and the actors who make them (politicians, parties, etc.). The contribution of this paper is twofold. First, we document and release DebateNet2.0 along with its companion R package, mardyR, guiding the reader through the practical and conceptual issues related to the annotation of policy debates in newspapers. Second, we outline and apply a Discourse Network Analysis (DNA) to DebateNet2.0, comparing two crucial moments of the policy debate on the 'refugee crisis': the migration flux through the Mediterranean in April/May and the one along the Balkan route in September/October. Besides the released resources and the case-study, our contribution is also methodological: we talk the reader through the steps from a newspaper article to a discourse network, demonstrating that there is not just one discourse network for the German migration debate, but multiple ones, depending on the topic of interest (political actors, policy fields, time spans).

Flor Miriam Plaza-del-Arco, Sercan Halat, Sebastian Padó et al.

The recognition of hate speech and offensive language (HOF) is commonly formulated as a classification task to decide if a text contains HOF. We investigate whether HOF detection can profit by taking into account the relationships between HOF and similar concepts: (a) HOF is related to sentiment analysis because hate speech is typically a negative statement and expresses a negative opinion; (b) it is related to emotion analysis, as expressed hate points to the author experiencing (or pretending to experience) anger while the addressees experience (or are intended to experience) fear. (c) Finally, one constituting element of HOF is the mention of a targeted person or group. On this basis, we hypothesize that HOF detection shows improvements when being modeled jointly with these concepts, in a multi-task learning setup. We base our experiments on existing data sets for each of these concepts (sentiment, emotion, target of HOF) and evaluate our models as a participant (as team IMS-SINAI) in the HASOC FIRE 2021 English Subtask 1A. Based on model-selection experiments in which we consider multiple available resources and submissions to the shared task, we find that the combination of the CrowdFlower emotion corpus, the SemEval 2016 Sentiment Corpus, and the OffensEval 2019 target detection data leads to an F1 =.79 in a multi-head multi-task learning model based on BERT, in comparison to .7895 of plain BERT. On the HASOC 2019 test data, this result is more substantial with an increase by 2pp in F1 and a considerable increase in recall. Across both data sets (2019, 2021), the recall is particularly increased for the class of HOF (6pp for the 2019 data and 3pp for the 2021 data), showing that MTL with emotion, sentiment, and target identification is an appropriate approach for early warning systems that might be deployed in social media platforms.

Sean Papay, Roman Klinger, Sebastian Padó

A major challenge in structured prediction is to represent the interdependencies within output structures. When outputs are structured as sequences, linear-chain conditional random fields (CRFs) are a widely used model class which can learn \textit{local} dependencies in the output. However, the CRF's Markov assumption makes it impossible for CRFs to represent distributions with \textit{nonlocal} dependencies, and standard CRFs are unable to respect nonlocal constraints of the data (such as global arity constraints on output labels). We present a generalization of CRFs that can enforce a broad class of constraints, including nonlocal ones, by specifying the space of possible output structures as a regular language $\mathcal{L}$. The resulting regular-constrained CRF (RegCCRF) has the same formal properties as a standard CRF, but assigns zero probability to all label sequences not in $\mathcal{L}$. Notably, RegCCRFs can incorporate their constraints during training, while related models only enforce constraints during decoding. We prove that constrained training is never worse than constrained decoding, and show empirically that it can be substantially better in practice. Additionally, we demonstrate a practical benefit on downstream tasks by incorporating a RegCCRF into a deep neural model for semantic role labeling, exceeding state-of-the-art results on a standard dataset.

Enrica Troiano, Sebastian Padó, Roman Klinger

When humans judge the affective content of texts, they also implicitly assess the correctness of such judgment, that is, their confidence. We hypothesize that people's (in)confidence that they performed well in an annotation task leads to (dis)agreements among each other. If this is true, confidence may serve as a diagnostic tool for systematic differences in annotations. To probe our assumption, we conduct a study on a subset of the Corpus of Contemporary American English, in which we ask raters to distinguish neutral sentences from emotion-bearing ones, while scoring the confidence of their answers. Confidence turns out to approximate inter-annotator disagreements. Further, we find that confidence is correlated to emotion intensity: perceiving stronger affect in text prompts annotators to more certain classification performances. This insight is relevant for modelling studies of intensity, as it opens the question wether automatic regressors or classifiers actually predict intensity, or rather human's self-perceived confidence.

Sean Papay, Roman Klinger, Sebastian Padó

Span identification (in short, span ID) tasks such as chunking, NER, or code-switching detection, ask models to identify and classify relevant spans in a text. Despite being a staple of NLP, and sharing a common structure, there is little insight on how these tasks' properties influence their difficulty, and thus little guidance on what model families work well on span ID tasks, and why. We analyze span ID tasks via performance prediction, estimating how well neural architectures do on different tasks. Our contributions are: (a) we identify key properties of span ID tasks that can inform performance prediction; (b) we carry out a large-scale experiment on English data, building a model to predict performance for unseen span ID tasks that can support architecture choices; (c), we investigate the parameters of the meta model, yielding new insights on how model and task properties interact to affect span ID performance. We find, e.g., that span frequency is especially important for LSTMs, and that CRFs help when spans are infrequent and boundaries non-distinctive.

Iris Hendrickx, Su Nam Kim, Zornitsa Kozareva et al.

In response to the continuing research interest in computational semantic analysis, we have proposed a new task for SemEval-2010: multi-way classification of mutually exclusive semantic relations between pairs of nominals. The task is designed to compare different approaches to the problem and to provide a standard testbed for future research. In this paper, we define the task, describe the creation of the datasets, and discuss the results of the participating 28 systems submitted by 10 teams.

Enrica Troiano, Sebastian Padó, Roman Klinger

Sentiment analysis has a range of corpora available across multiple languages. For emotion analysis, the situation is more limited, which hinders potential research on cross-lingual modeling and the development of predictive models for other languages. In this paper, we fill this gap for German by constructing deISEAR, a corpus designed in analogy to the well-established English ISEAR emotion dataset. Motivated by Scherer's appraisal theory, we implement a crowdsourcing experiment which consists of two steps. In step 1, participants create descriptions of emotional events for a given emotion. In step 2, five annotators assess the emotion expressed by the texts. We show that transferring an emotion classification model from the original English ISEAR to the German crowdsourced deISEAR via machine translation does not, on average, cause a performance drop.

Gemma Boleda, Sebastian Padó, Nghia The Pham et al.

Reference is a crucial property of language that allows us to connect linguistic expressions to the world. Modeling it requires handling both continuous and discrete aspects of meaning. Data-driven models excel at the former, but struggle with the latter, and the reverse is true for symbolic models. This paper (a) introduces a concrete referential task to test both aspects, called cross-modal entity tracking; (b) proposes a neural network architecture that uses external memory to build an entity library inspired in the DRSs of DRT, with a mechanism to dynamically introduce new referents or add information to referents that are already in the library. Our model shows promise: it beats traditional neural network architectures on the task. However, it is still outperformed by Memory Networks, another model with external memory.

Gemma Boleda, Sebastian Padó, Marco Baroni

One of the most basic functions of language is to refer to objects in a shared scene. Modeling reference with continuous representations is challenging because it requires individuation, i.e., tracking and distinguishing an arbitrary number of referents. We introduce a neural network model that, given a definite description and a set of objects represented by natural images, points to the intended object if the expression has a unique referent, or indicates a failure, if it does not. The model, directly trained on reference acts, is competitive with a pipeline manually engineered to perform the same task, both when referents are purely visual, and when they are characterized by a combination of visual and linguistic properties.