Papers by Noémie Elhadad
Learning to Revise References for Faithful Summarization (2022.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study shows that noisy reference summaries can be detrimental to model performance. |
| Approach: | They propose to selectively re-write unsupported reference sentences to better reflect source data. |
| Outcome: | The proposed method improves reference quality while retaining all data. |
Generating EDU Extracts for Plan-Guided Summary Re-Ranking (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to generate summary candidates for re-ranking produce redundant, and often low quality, content. |
| Approach: | They propose a method to generate candidates for re-ranking that addresses these issues by grounding each abstract on its own unique content plan and creating distinct plan-guided abstracts using a model's top beam. |
| Outcome: | The proposed method outperforms baseline decoding methods on CNN, NYT, and Xsum and shows that prompting GPT-3 to follow EDU plans outperformed sampling-based methods by 1.05 points. |
What’s in a Summary? Laying the Groundwork for Advances in Hospital-Course Summarization (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing methods to summarize clinical narratives are lacking. |
| Approach: | They propose to generate a paragraph that tells the story of a patient's hospitalization . they analyze a dataset of 109,000 hospitalizations and their corresponding summary proxy . |
| Outcome: | The proposed model is based on a dataset of 109,000 hospitalizations and their corresponding summary proxy. |
What are the Desired Characteristics of Calibration Sets? Identifying Correlates on Long Form Scientific Summarization (2023.acl-long)
Copied to clipboard
Griffin Adams, Bichlien Nguyen, Jake Smith, Yingce Xia, Shufang Xie, Anna Ostropolets, Budhaditya Deb, Yuan-Jyue Chen, Tristan Naumann, Noémie Elhadad
| Challenge: | Summarization models are trained to maximize the likelihood of a single reference (MLE) but little is known about why one setup is more effective than another . |
| Approach: | They add a calibration step which exposes a model to its own ranked outputs to improve relevance or contrasts positive and negative sets to improve faithfulness. |
| Outcome: | The proposed calibration step can unlock large gains in relevance or faithfulness. |