Papers by Stefan Dietze
Dissecting Paraphrases: The Impact of Prompt Syntax and supplementary Information on Knowledge Retrieval from Pretrained Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Pre-trained language models contain various kinds of knowledge. |
| Approach: | They designed a probe that allows comparison of 34 million distinct paraphrases that follow a unified meta-template enabling the controlled variation of syntax and semantics across arbitrary relations. |
| Outcome: | Extensive knowledge retrieval experiments show that prompts following clausal syntax have several desirable properties in comparison to appositive syntax. |
BERTweet’s TACO Fiesta: Contrasting Flavors On The Path Of Inference And Information-Driven Argument Mining On Twitter (2024.findings-naacl)
Copied to clipboard
| Challenge: | Argument mining is a challenging analytical task in the rich context of Twitter (now X). |
| Approach: | They propose to optimize the embeddings of the BERTweet transformer for argument mining on Twitter and broader generalization across topics. |
| Outcome: | The proposed approach improves classification and generalization across topics using a siamese network and a dataset. |
Limited Generalizability in Argument Mining: State-Of-The-Art Models Learn Datasets, Not Arguments (2025.acl-long)
Copied to clipboard
| Challenge: | Identifying arguments is a prerequisite for various tasks in automated discourse analysis. |
| Approach: | They evaluate four BERT-like transformers on 17 English sentence-level datasets . they find that they tend to rely on lexical shortcuts tied to content words . |
| Outcome: | The proposed models perform best on 17 English sentence-level datasets on common tasks, but their performance drops when applied to unseen datasets. |
TACO – Twitter Arguments from COnversations (2024.lrec-main)
Copied to clipboard
| Challenge: | Argument mining aims to identify the structural elements of arguments, denoted as information and inference, in online discourses. |
| Approach: | They propose to use Twitter Arguments to identify structural elements of arguments, denoted as information and inference, in a dataset that uses 1,814 tweets and an annotation framework that incorporates definitions from the Cambridge Dictionary to define and identify argument components. |
| Outcome: | The proposed dataset identifies arguments on Twitter and achieves an 85.06% macro F1 score in detecting arguments. |
GSAP-NER: A Novel Task, Corpus, and Baseline for Scholarly Entity Extraction Focused on Machine Learning Models and Datasets (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Named Entity Recognition (NER) models are crucial for academic writing . existing ground truth datasets do not treat fine-grained types like ML model and model architecture as separate entity types . |
| Approach: | They propose to annotate 100 full-text scientific publications and a first baseline model for 10 entity types centered around ML models and datasets. |
| Outcome: | The proposed model can be used to identify 10 entity types in scientific articles . existing models cannot recognize fine-grained models like ML models and model architecture . |