Deploy-a-sentiment-analysis

Udacity deep learning nanodegree project 5

Table of contents

  1. Software and Service used
  2. Project Inspiration
  3. File Descriptions
  4. Data Insights
  5. Licensing, Authors, and Acknowledgements

Software and service used

Python version 3.0. Amazon Web Service to train, evaluate and deploy the web app.

Project Inspiration

Get a sense of how to deploy a machine learning model using AWS Sagemaker, including upload train, validation and test data to S3; develope and train, validate and test RNN, then do the following as this picture indicate to deploy my web app that can judge a movie review is positive or negative.

screenshot

File Descriptions

Deploy_a_sentiment_analysis.ipynb : Jupyter notebook containing all the codes and results

report.html : notebbok in html format

train.py : model file

predict.py : predict fucntion file

index.html : web app

Insights

After deploy my web app, I generated few reviews and test it, the following are the results.

The intersting one is the last two, I said 'this movie is long and boring, but I like ...' and 'the only thing I like about this movie is it has the actress I like', the web app thinks both are NEGATIVE review and they do. Even tho I use plain and positive words like 'like', 'actress'. It feels like this web app understands my feeling, which feels funny in a way. Anyways, I only typed few reviews and there are room for improving, we could use LSTM RNN, etc....

screenshot screenshot screenshot screenshot

Licensing, Authors, Acknowledgements

Data: IMDb data

Authors: To the Notebook