Based on product reviews, we want to see how a review relates to ratings. In this project, a big topic called Text classification is implemented using two models. First, we implemented the LSTM model using Keras and train the binary classification. And the CNN model was implemented using Tensorflow to train and test multiple classification. We tested the post-learning using LSTM, which is a type of RNN known as a representative text processing model. Next, we implemented multiple text classification using CNN, which is mainly used for image processing. In conclusion, we have tested the possibility of how much performance can be expressed by text classification using RNN and CNN, which are representative algorithms of deep learning, and how much of the learning result, in this project . Especially, in order to process texts of CNN that performs image processing, the order of appearance of words and expressions is reflected in learning by preserving local information of sentences.
- Flask for web server
- Jupyter for running python program and leraning rnn/cnn
- tensorflow
- keras
$ python3 web.py
- In app/model/checkout, change a path
- paths of dataset
If you get dataset and model ipynb files and change the path of dataset, you can learn cnn or rnn model using dataset.
- Step1
- Step2
- Step3
- Result
- Step1
- Step2
- Result
- Understanding LSTM Networks
- How to Use Word Embedding Layers for Deep Learning with Keras
- What optimization methods work best for LSTMs?
- rnn vs cnn stackoverflow
- CNN text classification Implementation Github
- Convolutional Neural Networks for Sentence Classification
- kernel
- papers
- Dokyeong Kwon
- Seungwoo Park
- Taeseung Lee
https://www.slideshare.net/TaeseungLee1/prediction-of-amazon-review-rating-by-using-amazon-review