Papers by Dirk Padfield
MultiTurnCleanup: A Benchmark for Multi-Turn Spoken Conversational Transcript Cleanup (2023.emnlp-main)
Copied to clipboard
| Challenge: | Disfluency detection models focus on individual utterances, but discontinuities in spoken transcripts occur across multiple turns. |
| Approach: | They propose a multi-turn "cleanup task" to detect discontinuities in spoken conversations . they leverage two modeling approaches for experimental evaluation as benchmarks . |
| Outcome: | The proposed task detects "discontinuities" in spoken conversations that can be removed . the results are compared with existing methods and are expected to be validated in the future . |
Teaching BERT to Wait: Balancing Accuracy and Latency for Streaming Disfluency Detection (2022.naacl-main)
Copied to clipboard
| Challenge: | a recent study shows that current NLP models operate non-incrementally, causing unacceptable delays for the user. |
| Approach: | They propose a streaming BERT-based sequence tagging model that detects disfluencies in real-time . they train the model to decide whether to immediately output a prediction or wait for further context . |
| Outcome: | The proposed model produces accurate predictions sooner than baselines, with lower flicker . disfluencies hurt readability of ASR transcripts, erode model performance on downstream tasks . |
Barriers to Effective Evaluation of Simultaneous Interpretation (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing studies have relied on out-of-the-box machine translation metrics to evaluate interpretation data, but they do not account for human judgments of interpretation quality. |
| Approach: | They propose to use machine translation metrics to evaluate human interpretations to address potential barriers to disfluency, summarization, paraphrasing and segmentation. |
| Outcome: | The proposed model achieves better correlation with human judgments than state-of-the-art metrics. |
Residual Adapters for Parameter-Efficient ASR Adaptation to Atypical and Accented Speech (2021.emnlp-main)
Copied to clipboard
| Challenge: | Automatic Speech Recognition systems perform poorly on atypical speech and heavily accented speech. |
| Approach: | They add a residual adapter to the encoder layer to improve model adaptation . they show that the residual adapters update only a tiny fraction of the model parameters . |
| Outcome: | The proposed model fine-tuning improves performance on atypical and accented speech . the system can update only a tiny fraction of the model parameters . |