Papa_ting: Yelp Sentiment Analysis

Author: Ian-chin Wang We used yelp challenge dataset: https://www.yelp.com/dataset Due to the size of the data, we can not upload the data. We used review.json in the dataset as our raw data.

Flies

CS235ProjectReport.pdf: project final report source_code/: source code of the project models/: include source code of CNN and RNN models data/: include data for classification main.py: data preprocessing clustering/clustering.py: cluster the data from classification with K-means and tf-idf.(we have toy data for execute, and there are commands for executing in the comments of the file) clustering/dataset/test/: contain the toy data for clustering(contains positive and negative labeled reviews)

How to run

We present a Dockerfile for preprocessing and classification, for classification launch

docker run -v "$(pwd):/data_mining -t <name of container> <args>

Usage of our front-end ./main.py <preprocess/train/validate> -i <Input file or Directory> -t <cnn or lstm>

• Select preprocessing for converting the dataset into sentence, it take original review.json in Yelp dataset as input
• Select train for training, it read all file containing labeled data under a certain directory(From input), and save the model to the disk.