The goal of this project is to use Machine Learning to graph the mood (positive/negative) over a long piece of text (e.g. a book). In other words, it will graph sentiment over time.

Using this tool, you can turn a book (in this case Under The Dome by Steven King) into a map of emotion like this:

The red bars indicate negative mood, the green indicate positive mood, and white indicates neutral.

This project has three different components, so there are three main steps to using it.

The first step is to train a Machine Learning algorithm to determine the sentiment of a piece of text. You must download a training corpus for this (I recommend the one at http://help.sentiment140.com/for-students/). You will need to pick a location to save the trained classifier (I'll use /path/to/classifier):

$ go run train/*.go bayes /path/to/classifier /path/to/training.csv

This will take several minutes to run, and once it's done you will have a classifier.

The next step is to generate a CSV file with the sentiment of each sentence in the body of text you would like to graph. To do this, do the following:

$ go run plotcsv/*.go /path/to/classifier /path/to/text.txt /path/to/sentiments.csv

This will generate a file at /path/to/sentiments.csv containing sentiments for each sentence in the text file /path/to/text.txt.

Finally, to create a graphical image of the previously generated CSV file, you can do the following:

$ go run graph/*.go /path/to/sentiments.csv /path/to/graph.png heat

That will create a sentiment heat map out of the CSV file.