Papers by Katja Markert
Discrete Optimization for Unsupervised Sentence Summarization with Word-Level Extraction (2020.acl-main)
Copied to clipboard
| Challenge: | Sentence summarization systems that use latent space to reconstruct the source sentence are unwillingly exploited. |
| Approach: | They propose a method that uses language modeling and semantic similarity metrics to find a high-scoring summary. |
| Outcome: | The proposed method achieves state-of-the-art for unsupervised sentence summarization according to ROUGE scores. |
Abstractive Timeline Summarization (D19-54)
Copied to clipboard
| Challenge: | Prior approaches to TLS focus on extractive methods, which generate extractive timelines . a study with human judges shows that our abstractive system also produces output that is easy to read and understand. |
| Approach: | They propose an abstractive timeline summarization system that is unsupervised . their system outperforms extractive systems in terms of ROUGE scores . |
| Outcome: | The proposed system outperforms extractive systems in terms of ROUGE scores . it produces output that is easy to read and understand, the authors say . |
Distinguishing affixoid formations from compounds (C18-1)
Copied to clipboard
| Challenge: | affixoids are morphemes in between affids and free stems that have been associated with increased productivity and a bleached semantics but not empirically validated. |
| Approach: | They propose to use affixoids as morphemes in between affids and stems to test their classification in a subset of German words that includes many hapaxes. |
| Outcome: | The proposed morpheme can be classed as affixoid or non-affixoids with a best F1 score of 74% on a subset of German words that includes many hapaxes . |
Bias in News Summarization: Measures, Pitfalls and Corpora (2024.findings-acl)
Copied to clipboard
| Challenge: | Pretrained large language models can reproduce harmful social biases in constrained settings, such as summarization. |
| Approach: | They propose a method to generate input documents with carefully controlled demographic attributes and then apply it to a controlled setting. |
| Outcome: | The proposed method allows to generate input documents with carefully controlled demographic attributes while working with real-world input documents. |
Whose Facts Win? LLM Source Preferences under Knowledge Conflicts (2026.acl-long)
Copied to clipboard
| Challenge: | Existing studies on the role of the source of knowledge conflicts have not investigated the role . |
| Approach: | They propose a framework that reduces repetition bias by up to 79.2% while maintaining at least 72.5% of original preferences. |
| Outcome: | The proposed method reduces repetition bias by up to 79.2% while maintaining at least 72.5% of original preferences. |
How to Evaluate a Summarizer: Study Design and Statistical Analysis for Manual Linguistic Quality Evaluation (2021.eacl-main)
Copied to clipboard
| Challenge: | Current manual evaluation methods for text summarization have low correlation with human judgements on summary quality. |
| Approach: | They conduct two evaluation experiments on two aspects of summaries’ linguistic quality (coherence and repetitiveness) they find that study parameters such as the total number of annotators and distribution of annotes to annotation items are often not fully reported. |
| Outcome: | The proposed methods can inflate type I errors up to eight-fold and the overall number of annotators can have a strong impact on study power. |
Dataset Reproducibility and IR Methods in Timeline Summarization (2020.lrec-1)
Copied to clipboard
| Challenge: | Timeline summarization (TLS) generates a dated overview of real-world events based on event-specific corpora. |
| Approach: | They propose to use IR methods to construct event-specific corpora from a newsroom dataset . they advocate for integrating IR into the development of TLS systems . |
| Outcome: | The proposed method is not reproducible at different search times and uses components that are not always available for large news corpus. |
An analysis of language models for metaphor recognition (2020.coling-main)
Copied to clipboard
| Challenge: | Metaphor recognition systems that are based on language models perform substantially worse on unconventional metaphors than on conventional ones. |
| Approach: | They conduct a linguistic analysis of recent metaphor recognition systems based on language models and a variant of BERT language models to examine their performance. |
| Outcome: | The proposed systems show that they can recognise unseen words if synonyms or morphological variations have been seen before, leading to enhanced generalisation beyond word sense disambiguation. |
With a Little Push, NLI Models can Robustly and Efficiently Predict Faithfulness (2023.acl-short)
Copied to clipboard
| Challenge: | Conditional language models generate unfaithful output that is not supported by their input . this jeopardizes trust in real-world applications, raising a need for automatic faithfulness metrics. |
| Approach: | They propose to augment conditional language models with robust inference procedures to improve faithfulness. |
| Outcome: | The proposed approach outperforms existing models on the TRUE benchmark. |
Biographically Relevant Tweets – a New Dataset, Linguistic Analysis and Classification Experiments (2022.coling-1)
Copied to clipboard
| Challenge: | Unlike previous work, we do not restrict biographical relevance to a small fixed set of pre-defined relations. |
| Approach: | They propose a dataset comprising tweets for the novel task of detecting biographically relevant utterances. |
| Outcome: | The proposed dataset focuses on biographical information on ordinary users of Twitter. |
How to Find Strong Summary Coherence Measures? A Toolbox and a Comparative Study for Summary Coherence Measure Evaluation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods to evaluate summary coherence are often evaluated using disparate datasets and metrics. |
| Approach: | They propose to use automatic evaluation to evaluate coherence of summaries by selecting high-scoring candidates. |
| Outcome: | The proposed methods show that they can perform better on an even playing field. |
Doctor Who? Framing Through Names and Titles in German (2020.lrec-1)
Copied to clipboard
| Challenge: | Entity framing is the selection of aspects of an entity to promote a particular viewpoint towards that entity. |
| Approach: | They investigate entity framing of political figures through the use of names and titles in German online discourse. |
| Outcome: | The proposed method improves existing studies on German political discourse . it shows that the formality of naming correlates positively with stance in the tweets . |
The Chinese Causative-Passive Homonymy Disambiguation: an adversarial Dataset for NLI and a Probing Task (2022.lrec-1)
Copied to clipboard
| Challenge: | Recent research questions whether these models really understand the meaning of natural language. |
| Approach: | They propose to transform the disambiguation of causative-passive homonymy (CPH) to a challenging natural language inference task using a pretrained transformer model RoBERTa. |
| Outcome: | The pretrained model RoBERTa performs poorly on the CANLI dataset . the model's internal representation of CPH is not captured in the model . |
Context in Informational Bias Detection (2020.coling-main)
Copied to clipboard
| Challenge: | Informational bias is conveyed through sentences or clauses that provide tangential, speculative or background information that can sway readers’ opinions towards entities. |
| Approach: | They explore four kinds of context for informational bias in English news articles . integrating event context improves classification performance over a strong baseline . |
| Outcome: | The best-performing model outperforms the baseline on longer sentences and sentences from politically centrist articles. |