Here, a Multi-Label classification study is carried out. A design has been made with the Bidirectional Deep Learning model. Implementation made with Keras. In the study, it is also ensured that the relevant word is highlighted with the attention mechanism.
- keras framework was used, RNN application was processed.
- it is aimed to show which or which of the 6 different tags that sentence covers in any sentence.
- the attention mechanism was established and put into practice. "Hierarchial Attention Network" was used to find important words affecting the label. In this way, important words in the text have become to be identified.
Dataset: https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/data