Beyond Static Models and Test Sets: Benchmarking the Potential of Pre-trained Models Across Tasks and Languages
This addresses the problem of incomplete benchmarking in multilingual NLP for researchers and practitioners, offering a more efficient and cost-effective evaluation method, though it is incremental as it builds on existing performance prediction work.
The paper tackles the unreliable evaluation of Massively Multilingual Language Models (MMLMs) due to limited linguistic diversity in benchmarks by proposing performance prediction methods based on data and language typology features. The result shows that these methods provide reliable performance estimates comparable to translation-based approaches, without additional translation or evaluation costs.
Although recent Massively Multilingual Language Models (MMLMs) like mBERT and XLMR support around 100 languages, most existing multilingual NLP benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity. We argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of MMLMs across the linguistic landscape. We propose that the recent work done in Performance Prediction for NLP tasks can serve as a potential solution in fixing benchmarking in Multilingual NLP by utilizing features related to data and language typology to estimate the performance of an MMLM on different languages. We compare performance prediction with translating test data with a case study on four different multilingual datasets, and observe that these methods can provide reliable estimates of the performance that are often on-par with the translation based approaches, without the need for any additional translation as well as evaluation costs.