GLEU Without Tuning
arXiv:1605.02592
Abstract
The GLEU metric was proposed for evaluating grammatical error corrections using n-gram overlap with a set of reference sentences, as opposed to precision/recall of specific annotated errors (Napoles et al., 2015). This paper describes improvements made to the GLEU metric that address problems that arise when using an increasing number of reference sets. Unlike the originally presented metric, the modified metric does not require tuning. We recommend that this version be used instead of the original version.
Cited by in corpus (10)
- Attention-based Encoder-Decoder Networks for Spelling and Grammatical Error Correction
- Weakly Supervised Grammatical Error Correction using Iterative Decoding
- An Empirical Study of Incorporating Pseudo Data into Grammatical Error Correction
- Shifts: A Dataset of Real Distributional Shift Across Multiple Large-Scale Tasks
- Parallel Iterative Edit Models for Local Sequence Transduction
- Automated Query Reformulation for Efficient Search based on Query Logs From Stack Overflow
- CrowdSpeech and VoxDIY: Benchmark Datasets for Crowdsourced Audio Transcription
- A Self-Refinement Strategy for Noise Reduction in Grammatical Error Correction
- Comparison of Grammatical Error Correction Using Back-Translation Models
- Corpora Generation for Grammatical Error Correction