2020/10/07 by Eleftheria Briakou, Marine Carpuat, Briakou, Eleftheria +1
Computer Science · #Natural Language Processing Techniques #Topic Modeling #Text Readability and Simplification
paper · pdf · doi:10.48550/arxiv.2010.03662
Detecting fine-grained differences in content conveyed in different languages\nmatters for cross-lingual NLP and multilingual corpora analysis, but it is a\nchallenging machine learning problem since annotation is expensive and hard to\nscale. This work improves the prediction and annotation of fine-grained\nsemantic divergences. We introduce a training strategy for multilingual BERT\nmodels by learning to rank synthetic divergent examples of varying granularity.\nWe evaluate our models on the Rationalized English-French Semantic Divergences,\na new dataset released with this work, consisting of English-French\nsentence-pairs annotated with semantic divergence classes and token-level\nrationales. Learning to rank helps detect fine-grained sentence-level\ndivergences more accurately than a strong sentence-level similarity model,\nwhile token-level predictions have the potential of further distinguishing\nbetween coarse and fine-grained divergences.\n