Predicting Performance for Natural Language Processing Tasks
A blog accompanying our ACL 2020 paper (with Mengzhou Xia, Ruochen Xu, Yiming Yand, and Graham Neubig). We use previously published results to get an estimation of the performance on a new NLP experiment, before running it! We describe NLP experiments with a set of interpretable features, and train prediction models which are actually pretty good at producing such estimations, and show how they can be useful in various scenarios.
[read more]
[paper]
Should All Cross-Lingual Embeddings Speak English?
A blog accompanying our ACL 2020 paper. The answer to the above question is "No, duh!", backed by results with multiple Lexicon Induction methods. We also provide more than 4000 dictionaries for evaluating LI methods on non-English language pairs.
[read more]
[paper]
[dataset]
A note on evaluating multilingual benchmarks
Be careful when reporting averages for multilingual benchmarks, especially if making claims about multilinguality. In addition, averaging by language family can provide additional insights.
[read more]