Please enable JavaScript.

Coggle requires JavaScript to display documents.

Introduction (Motivation (Objectives (Related Work (Methodology (Problem…

- - - - Text Representation
        
        Based on Distributional Hypothesis
        
        Advanced embedding method which is transferred from the model learned on different high-level language understanding tasks.
      - Comparison of Semantic Textual Similarity on different embedding
      - Methodology
        
        Overview
        
        Dataset and data operations
        
        Classifier
        
        Problem Statement
        
        Experiment 1
        
        What is the effect of SDA on text clas-sification?
        
        Experiment 2
        
        Experiment 3
        
        Conclusion
        
        Future works
- - - - Solution
        
        Collecting annotated data
        
        Data augmentation
        
        The aims of DA
        
        Reducing overfitting
        
        Improving performance
        
        DA on Computer Vision
        
        DA on Linguistic Data
        
        Replacing synonym
        
        Limitation of text data augmentation
        
        If simply transfer operations from images to texts.
        
        Augmentation from embeddings
      - Limitation of Generated-based Data Augmentation
        
        Sampled-based Data Augmentation
        
        The aims of Sampled-based data augmentation is to identify appropriate samples as good candidate samples through a criterion
        
        WDA: Web Data Augmentation
        
        advantage