###### tags: Paper Reading
# Hierarchical Neural Networks for Sequential Sentence Classification
## Outline
    This paper is talking about Sequential Sentence Classification in Medical Scientific Abstracts.
## Introduction/Motivation
    In the traditional sentence classification problem, we always focus on classifying single sentence rather than taking rhetorical sections into consideration. So the paper presented a structure base on artificial neural network which is called hierarchical sequential labeling network (HSLN). HSLN make use of the contextual information within surrounding sentences to help classify the current sentence.
## Model
1.    In the first, puts sentence of words into Token Embedding Layer to **represent sentences**.
1.    Second, puts the sentence of embedding into CNN or RNN to **encoding embedding** and take the use of attention mechanism at the same time.
1.     After get the encoding of every sentence, puts them into a bi-directional LSTM to **find the contextual information about the abstracts**.
1.     In the end, puts all the output from LSTM into forward layer and CRF layer to get **final gold label**.
## Experiment
    section5
## Conclusion
   Most of the layer mentioned by the paper are also what people will do about the text classification problem now. This paper provide the basic idea and knowledge about Text Categorization for me. After reading this paper, I gain the ability about Text Analysis.