Automatic Construction of Evaluation Suites for Natural Language Generation Datasets
This addresses the issue of biased and incomplete model evaluation in NLP, particularly for underrepresented groups and rare language phenomena, though it is incremental as it builds on existing challenge set ideas.
The paper tackles the problem of oversimplified evaluation in natural language generation by developing a framework to automatically generate challenge sets that assess specific model capabilities, resulting in an 80-challenge evaluation suite for the GEM benchmark that reveals limitations of current models.
Machine learning approaches applied to NLP are often evaluated by summarizing their performance in a single number, for example accuracy. Since most test sets are constructed as an i.i.d. sample from the overall data, this approach overly simplifies the complexity of language and encourages overfitting to the head of the data distribution. As such, rare language phenomena or text about underrepresented groups are not equally included in the evaluation. To encourage more in-depth model analyses, researchers have proposed the use of multiple test sets, also called challenge sets, that assess specific capabilities of a model. In this paper, we develop a framework based on this idea which is able to generate controlled perturbations and identify subsets in text-to-scalar, text-to-text, or data-to-text settings. By applying this framework to the GEM generation benchmark, we propose an evaluation suite made of 80 challenge sets, demonstrate the kinds of analyses that it enables and shed light onto the limits of current generation models.