Towards More Fine-grained and Reliable NLP Performance Prediction
arXiv:2102.05486
Abstract
Performance prediction, the task of estimating a system's performance without performing experiments, allows us to reduce the experimental burden caused by the combinatorial explosion of different datasets, languages, tasks, and models. In this paper, we make two contributions to improving performance prediction for NLP tasks. First, we examine performance predictors not only for holistic measures of accuracy like F1 or BLEU but also fine-grained performance measures such as accuracy over individual classes of examples. Second, we propose methods to understand the reliability of a performance prediction model from two angles: confidence intervals and calibration. We perform an analysis of four types of NLP tasks, and both demonstrate the feasibility of fine-grained performance prediction and the necessity to perform reliability analysis for performance prediction methods in the future. We make our code publicly available: \url{https://github.com/neulab/Reliable-NLPPP}
arXiv commentAccepted by EACL 2021
References in corpus (6)
- Natural Language Processing (almost) from Scratch
- Visualizing and Understanding Recurrent Networks
- A Convolutional Neural Network for Modelling Sentences
- A Constructive Prediction of the Generalization Error Across Scales
- compare-mt: A Tool for Holistic Comparison of Language Generation Systems
- GRN: Gated Relation Network to Enhance Convolutional Neural Network for Named Entity Recognition