Improving Named Entity Recognition in Telephone Conversations via Effective Active Learning with Human in the Loop
This addresses the challenge of annotation errors in noisy data for NER in telephone conversations, offering a practical solution to improve model accuracy with minimal re-annotation effort, though it is incremental in nature.
The paper tackles the problem of poor model performance in Named Entity Recognition due to annotation errors in noisy telephone transcription data by proposing an active learning framework that identifies likely erroneous samples for re-annotation. The result is that re-annotating only 6% of training instances improves the F1 score for a certain entity type by about 25%.
Telephone transcription data can be very noisy due to speech recognition errors, disfluencies, etc. Not only that annotating such data is very challenging for the annotators, but also such data may have lots of annotation errors even after the annotation job is completed, resulting in a very poor model performance. In this paper, we present an active learning framework that leverages human in the loop learning to identify data samples from the annotated dataset for re-annotation that are more likely to contain annotation errors. In this way, we largely reduce the need for data re-annotation for the whole dataset. We conduct extensive experiments with our proposed approach for Named Entity Recognition and observe that by re-annotating only about 6% training instances out of the whole dataset, the F1 score for a certain entity type can be significantly improved by about 25%.