Papers by Thomas Schaaf

5 papers
Effective Convolutional Attention Network for Multi-label Clinical Document Classification (2021.emnlp-main)

Copied to clipboard

Challenge: a large number of medical encounters need to be coded everyday due to long document sets and large label set.
Approach: They propose a convolutional attention network for multi-label document classification problem . they use convolution-based encoders and convolution networks to aggregate information across documents .
Outcome: The proposed model outperforms prior best model and multilingual Transformer model on a widely used dataset in the medical domain.
Annotate the Way You Think: An Incremental Note Generation Framework for the Summarization of Medical Conversations (2024.lrec-main)

Copied to clipboard

Challenge: Existing datasets for summarization of medical conversations are limited to conversation-summary pairs . a novel annotation framework is proposed to capture the summarizing process via an annotation task .
Approach: They propose an incremental note generation framework that captures the human summarization process via an annotation task by instructing annotators to first incrementally create a draft note and polish it into a reference note.
Outcome: The proposed framework shows that the human summarization process is much more efficient and accurate than the current method.
Comparing Two Model Designs for Clinical Note Generation; Is an LLM a Useful Evaluator of Consistency? (2024.findings-naacl)

Copied to clipboard

Challenge: a clinical note is a document that documents a doctor's interaction with a patient . authors show that LLMs can be used to measure quality indicators .
Approach: They analyze two different approaches to generate different sections of a SOAP note . they use PEGASUS-X Transformer models to examine note consistency .
Outcome: The proposed approach leads to similar ROUGE values and no difference in Factuality metric . human reviewers perform the same tasks with roughly the same agreement as the LLMs .
Leveraging Pretrained Models for Automatic Summarization of Doctor-Patient Conversations (2021.findings-emnlp)

Copied to clipboard

Challenge: Using pretrained transformer models for automatically summarizing doctor-patient conversations presents challenges . limited training data, domain shift, long and noisy transcripts, and high target summary variability are challenges compared to human annotators.
Approach: They propose a method for fine-tuning pretrained transformer models for automatically summarizing doctor-patient conversations directly from transcripts.
Outcome: The proposed method surpasses the performance of an average human annotator and the quality of previous published work for the task.
Posterior Calibrated Training on Sentence Classification Tasks (2020.acl-main)

Copied to clipboard

Challenge: Existing methods for posterior calibration have been used to correct poorly calibrated posterior probabilities.
Approach: They propose a posterior calibration procedure that optimizes posterior probability distributions while minimizing calibration errors.
Outcome: The proposed procedure reduces calibration error and improves performance on both objectives.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations