Papers by Lun-Wei Ku
HonestBait: Forward References for Attractive but Faithful Headline Generation (2023.findings-acl)
Copied to clipboard
| Challenge: | Current approaches to generating attractive headlines often learn directly from data based on clicks and views . clickbait models fail to reveal how much interest is raised by the writing style and how much is due to the event or topic itself . |
| Approach: | They propose a framework for generating headlines using forward references . they use a dataset containing pairs of fake news and verified news . |
| Outcome: | The proposed framework yields more attractive headlines while maintaining high veracity . the framework is based on a dataset containing fake news with verified news . |
Beyond Evidence: Belief-Chain Conditioning for Persuasive Misinformation Debunking Explanation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to misinformation correction focus on relying on audience beliefs to generate factually accurate responses and to engage with users' mental states. |
| Approach: | They construct large language models with cognitive chains and use them to model their outputs on beliefs that engage with users' mental states. |
| Outcome: | The proposed model improves explanation quality for audiences with misinformation-aligned beliefs by incorporating believers’ chains into the model. |
Label-Aware Hyperbolic Embeddings for Fine-grained Emotion Classification (2023.acl-long)
Copied to clipboard
| Challenge: | Existing models only address text classification problem in the euclidean space, which is not optimal . e.g., fear and terrified labels may not be differentiated in such space, harming performance . |
| Approach: | They propose a framework that can integrate hyperbolic embeddings to improve the task . they learn label embeddements in the hyperbolical space and then add them to the framework . |
| Outcome: | The proposed framework improves fine-grained emotion classification on two benchmark datasets with 3% improvement over previous state-of-the-art models. |
Assessing the Helpfulness of Learning Materials with Inference-Based Learner-Like Agent (2020.emnlp-main)
Copied to clipboard
| Challenge: | Prior work uses hand-crafted scores to recommend sentences but has difficulty adopting such scores to all the near-synonyms as near-near-sonyms differ in various ways. |
| Approach: | They propose an inference-based learner-like agent to mimic learner behavior and identify good learning materials by examining the agent's performance. |
| Outcome: | The proposed agent achieves the best performance in fill-in-the-blank and good example sentence selection tasks. |
LLM-in-the-loop: Leveraging Large Language Model for Thematic Analysis (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent research shows that large language models can replicate human-like behavior in various tasks. |
| Approach: | They propose a framework for human-LLM collaboration to conduct TA with in-context learning (ICL) they propose to use survey data to frame discussions with an LLM to generate a final codebook for TA. |
| Outcome: | The proposed framework outperforms crowd workers on text-annotation tasks and yields similar coding quality to that of human coders but reduces TA’s labor and time demands. |
Learning to Rank Visual Stories From Human Ranking Data (2022.acl-long)
Copied to clipboard
| Challenge: | Existing studies on visual storytelling (VIST) use automated evaluation metrics for text generation. |
| Approach: | They develop a Vrank metric that repurposes human evaluation results for automatic evaluation. |
| Outcome: | The proposed model is more accurate than existing metrics and is generalizable to textual stories. |
UHop: An Unrestricted-Hop Relation Extraction Framework for Knowledge-Based Question Answering (N19-1)
Copied to clipboard
| Challenge: | Existing work restricts search from one entity to another to the maximum number of hops . a knowledge graph is a powerful graph structure that encodes knowledge to save and organize it . |
| Approach: | They propose an unrestricted-hop framework which relaxes the restriction by using a transition-based search framework. |
| Outcome: | The proposed framework performs well with state-of-the-art models and is competitive without exhaustive searches. |
EmotionLines: An Emotion Corpus of Multi-Party Conversations (L18-1)
Copied to clipboard
| Challenge: | Emotion is a critical characteristic to distinguish people from machines. |
| Approach: | They propose a dataset with emotions labeling on all utterances in each dialogue . they use Friends TV scripts and Facebook messenger dialogues to collect the data . |
| Outcome: | The proposed dataset is the first with emotions labeling on all utterances in each dialogue based on their textual content. |
Multi-VQG: Generating Engaging Questions for Multiple Images (2022.emnlp-main)
Copied to clipboard
| Challenge: | Traditional visual question generation (VQG) focuses on single images, resulting in a limited ability to comprehend time-series information of the underlying event. |
| Approach: | They propose to generate engaging questions from multiple images using a visual question generation dataset and establish a series of baselines. |
| Outcome: | The proposed model builds stories behind the image sequence to allow for creativity and experience sharing and hence draw attention to downstream applications. |
CoachMe: Decoding Sport Elements with a Reference-Based Coaching Instruction Generation Model (2025.acl-long)
Copied to clipboard
Wei-Hsin Yeh, Yu-An Su, Chih-Ning Chen, Yi-Hsueh Lin, Calvin Ku, Wenhsin Chiu, Min-Chun Hu, Lun-Wei Ku
| Challenge: | Existing multimodal models for motion related tasks have shown significant progress. |
| Approach: | They propose a reference-based model that analyzes the differences between a learner’s motion and a physical reference under temporal and physical aspects. |
| Outcome: | The proposed model outperforms GPT-4o on figure skating and boxing by 31.6% and 58.3% respectively. |
Plot and Rework: Modeling Storylines for Visual Storytelling (2021.findings-acl)
Copied to clipboard
| Challenge: | Automated visual storytelling models do not make extensive use of external knowledge and iterative generation when attempting to create stories. |
| Approach: | They propose a framework that uses an image sequence as a story graph to create a coherent story. |
| Outcome: | The proposed framework produces stories superior in diversity, coherence, and humanness . it uses plotting and reworking to improve the model's performance, the authors say . |
Stretch-VST: Getting Flexible With Visual Stories (2021.acl-demo)
Copied to clipboard
| Challenge: | Existing visual storytelling models produce stories with fixed lengths of five sentences and the fix-length stories carry limited details and provide ambiguous textual information to the readers. |
| Approach: | They propose to “stretch” visual storytelling frameworks by adding appropriate knowledge to the model to generate long stories. |
| Outcome: | The proposed framework provides better focus and detail when long stories are generated without deteriorating the quality. |
Is Explanation the Cure? Misinformation Mitigation in the Short Term and Long Term (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Using natural language processing (NLP), there is an ongoing shift towards NLPbased solutions such as fake news detection and generation of fact-checked, counterfactual explanations. |
| Approach: | They compare the effectiveness of a warning label and state-of-the-art counterfactual explanations generated by natural language generation (GPT4) models in debunking misinformation. |
| Outcome: | The proposed explanations significantly decrease participants’ self-reported belief in fake claims for the short-term and long-term. |
Beyond Fair Pay: Ethical Implications of NLP Crowdsourcing (2021.naacl-main)
Copied to clipboard
| Challenge: | Ethical considerations regarding the use of crowdworkers are limited to labor conditions . the Final Rule did not anticipate the use online crowdsourcing platforms for data collection . |
| Approach: | They propose to reopen discussion regarding ethical use of crowdworkers in NLP research . they propose to use online crowdsourcing platforms to evaluate risk of harm . |
| Outcome: | The proposed study identifies common scenarios where crowdworkers performing NLP tasks are at risk of harm. |
Lying Through One’s Teeth: A Study on Verbal Leakage Cues (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on verbal leakage cues do not address their impact on models' validity. |
| Approach: | They propose to use LIWC to show verbal leakage cues in lie detection datasets to understand their effect on data collection and examine their validity. |
| Outcome: | The proposed models with more strong verbal leakage cue categories perform better than models trained on a dataset with only a greater number of strong cues. |
Reactive Supervision: A New Method for Collecting Sarcasm Data (2020.emnlp-main)
Copied to clipboard
| Challenge: | sarcasm detection requires large amounts of labeled data, with a high cost and noisy labels. |
| Approach: | They propose a method that uses the dynamics of online conversations to collect sarcasm data. |
| Outcome: | The proposed method can be adapted to other affective computing domains, opening up new research opportunities. |
Expert Calibration Lens for Pruning Mixture of Experts (2026.acl-demo)
Copied to clipboard
| Challenge: | Expert pruning is a practical deployment technique for Mixture-of-Experts models . but its success depends heavily on the calibration set used for pruning . |
| Approach: | They propose a calibration tool that compares expert activations across datasets to predict calibration perturbations without running expensive pruning procedures. |
| Outcome: | The proposed system compares expert activations across datasets to predict calibration perturbations without running expensive pruning procedures. |
Happy Dance, Slow Clap: Using Reaction GIFs to Predict Induced Affect on Twitter (2021.acl-short)
Copied to clipboard
| Challenge: | Existing methods for labeling emotions in text are limited, but they can be useful for many tasks. |
| Approach: | They propose a method to collect texts with induced emotion and induced sentiment labels. |
| Outcome: | The proposed method can augment the data with induced emotion and induced sentiment labels. |
Do Large Multimodal Models Solve Caption Generation for Scientific Figures? Lessons Learned from SciCap Challenge 2023 (2026.tacl-1)
Copied to clipboard
Ting-Yao Hsu, Yi-Li Hsu, Shaurya Rohatgi, Chieh-Yang Huang, Ho Yin Sam Ng, Ryan Rossi, Sungchul Kim, Tong Yu, Lun-Wei Ku, Clyde Lee Giles, Ting-Hao Huang
| Challenge: | SciCap dataset launched in 2021 aims to generate high-quality captions for scientific figures. |
| Approach: | They propose to use the SciCap dataset to develop models for captioning diverse figure types across various academic fields. |
| Outcome: | The proposed models showed impressive performance on the SciCap dataset and in various vision-and-language tasks. |
Location-Aware Visual Question Generation with Lightweight Models (2023.emnlp-main)
Copied to clipboard
Nicholas Suwono, Justin Chen, Tun Hung, Ting-Hao Huang, I-Bin Liao, Yung-Hui Li, Lun-Wei Ku, Shao-Hua Sun
| Challenge: | a novel task aims to generate engaging questions from location-aware information . a lightweight model can be used to generate such questions . |
| Approach: | They propose a task to generate engaging questions from location-aware data . they represent location-based information with surrounding images and a GPS coordinate . |
| Outcome: | The proposed method outperforms baselines regarding human evaluation and evaluation metrics. |
Enhancing Perception: Refining Explanations of News Claims with LLM Conversations (2024.findings-naacl)
Copied to clipboard
| Challenge: | a new framework for Large Language Models (LLMs) streamlines the task of crafting explanations for fake news . a study compared refinement conversations between human and LLMs to enhance the effectiveness of LLM explanations . |
| Approach: | They propose a framework for Large Language Models to streamline the task of crafting fake news explanations. |
| Outcome: | The proposed framework enhances the process of crafting explanations for fake news claims through conversational refinement. |