PECO: Examining Single Sentence Label Leakage in Natural Language Inference Datasets through Progressive Evaluation of Cluster Outliers.
Michael SaxonXinyi WangWenda XuWilliam Yang WangPublished in: EACL (2023)
Keyphrases
- natural language
- natural language processing
- data points
- semantic representation
- information extraction
- bayesian networks
- clustering algorithm
- machine translation
- natural language interface
- knowledge representation
- missing data
- semantic role labeling
- semantic analysis
- data sets
- inference process
- natural language sentences
- clustering approaches
- semantic interpretation
- arbitrary shape
- cluster analysis
- multi label
- benchmark datasets
- feature space
- training data
- machine learning