Looking Beyond Label Noise: Shifted Label Distribution Matters in Distantly Supervised Relation Extraction.
Qinyuan YeLiyuan LiuMaosen ZhangXiang RenPublished in: EMNLP/IJCNLP (1) (2019)
Keyphrases
- label noise
- relation extraction
- active learning
- automatic extraction
- information extraction
- named entities
- noise tolerant
- multiple instance learning
- domain specific
- noise model
- named entity recognition
- semantic relations
- training data
- semi supervised
- supervised learning
- question answering
- semantic features
- gaussian distribution
- machine learning
- learning algorithm
- unsupervised learning
- data sets
- multi label
- class labels
- semi supervised learning
- higher order
- image retrieval
- training set
- reinforcement learning