Are Pretrained Transformers Robust in Intent Classification? A Missing Ingredient in Evaluation of Out-of-Scope Intent Detection.
Jianguo ZhangKazuma HashimotoYao WanYe LiuCaiming XiongPhilip S. YuPublished in: CoRR (2021)
Keyphrases
- robust detection
- pattern recognition
- preprocessing
- support vector machine
- machine learning
- classification algorithm
- feature space
- feature vectors
- pattern classification
- feature extraction and classification
- text classification
- object detection
- classification accuracy
- support vector
- digital mammograms
- decision trees
- unsupervised learning
- automatic detection
- support vector machine svm
- classification systems
- detection rate
- classification method
- microcalcification clusters
- data sets
- cost sensitive
- evaluation method
- false alarms
- neyman pearson
- driver assistance systems
- partial occlusion
- benchmark datasets
- machine learning algorithms
- detection method
- detection algorithm
- computationally efficient
- anomaly detection
- multi class
- face recognition
- feature selection
- computer vision