LSTM-based encoders perform better than dilated CNNs, Transformers, and BERT in this task . Domain-specific WORD2VEC embeddings outperform generic pre-trainedGLOVE . Morpho-syntactic features in the form of POS tag and tokenshape embeddeddings do not improve performance . Several observations contradict previous work on contract element extraction and sequence labeling tasks .

Author(s) : Ilias Chalkidis, Manos Fergadiotis, Prodromos Malakasiotis, Ion Androutsopoulos

Links : PDF - Abstract

Code :

Keywords : extraction - element - contract - - performance -

Leave a Reply

Your email address will not be published. Required fields are marked *