Robust Named Entity Recognition with Truecasing Pretraining
This addresses robustness issues in NER for noisy text, which is an incremental improvement over existing methods.
The paper tackles the problem of named entity recognition (NER) systems performing poorly on noisy data with uncertain capitalization by introducing a truecasing pretraining objective. It shows that this method improves performance on uncased text, achieving a new state of the art on the WNUT17 dataset.
Although modern named entity recognition (NER) systems show impressive performance on standard datasets, they perform poorly when presented with noisy data. In particular, capitalization is a strong signal for entities in many languages, and even state of the art models overfit to this feature, with drastically lower performance on uncapitalized text. In this work, we address the problem of robustness of NER systems in data with noisy or uncertain casing, using a pretraining objective that predicts casing in text, or a truecaser, leveraging unlabeled data. The pretrained truecaser is combined with a standard BiLSTM-CRF model for NER by appending output distributions to character embeddings. In experiments over several datasets of varying domain and casing quality, we show that our new model improves performance in uncased text, even adding value to uncased BERT embeddings. Our method achieves a new state of the art on the WNUT17 shared task dataset.