Cross-Lingual Event Detection via Optimized Adversarial Training

Anonymous

Abstract: In this work, we focus on Cross-Lingual Event Detection where a model is trained on data from a source language but its performance is evaluated on data from a second, target, language. Most recent works in this area have harnessed the language-invariant qualities displayed by pre-trained Multi-lingual Language Models. Their performance, however, reveals there is room for improvement as they mishandle delicate cross-lingual instances. We employ Adversarial Language Adaptation to train a Language Discriminator to discern between the source and target languages using unlabeled data. The discriminator is trained in an adversarial manner so that the encoder learns to produce refined, language-invariant representations that lead to improved performance. More importantly, we optimize the adversarial training by only presenting the discriminator with the most informative samples. We base our intuition about what makes a sample informative on two disparate metrics: sample similarity and event presence. Thus, we propose using Optimal Transport as a solution to naturally combine these two distinct information sources into the selection process. Extensive experiments on 8 different language pairs, using 4 languages from unrelated families, show the flexibility and effectiveness of our model that achieves new state-of-the-art results.

Paper Type: long