Automator

Naive Bayes on top of a Dropbox Datastore. Put together by Ryhan.

Setup

1. Authenticate your user with Dropbox

The Dropbox Datastore JavaScript Tutorial has better documentation on this, but the gist is

var client = new Dropbox.Client({key: APP_KEY});

// Try to finish OAuth authorization.
client.authenticate({interactive: false});

2. Open a datastore and use it to create an Automator

// Our to-be Automator object
var myClassifier;

if (client.isAuthenticated()) {
  var datastoreManager = client.getDatastoreManager();
  datastoreManager.openDefaultDatastore(function (error, datastore) {

    // Make an Automator object by passing in your datastore.
    myClassifier = new Automator(datastore);
  }
}

Usage

Train your classifier

Once you have created your classifier, you can train it by supplying a string of space delimited lowercase words and a category (classification) name.

myClassifier.train("some lowercase space delimited text", "spam");
myClassifier.train("man I really love bagels", "not spam");

Classify something

Once you have sufficient examples, you can start to classify some text.

myClassifier.classify("what are bagels like");

The above will return

{
  category: "not spam", // returns a classification or "unknown"
  reason: ["bagels"]    // returns an array of features that support this classification
  confidence: 0.5       // some number from 0 to 1
}

Preserve / Restore / Merge State

Oftentimes you want to start users off with some sort of default state.

You can check if the model has been trained at all by calling

myClassifier.hasModel() // Returns a boolean

You can destroy an existing model by calling

myClassifier.clearModel()

You can export your classifier's current state by calling

var state = myClassifier.toJSON();

and similarly merge/restore state by calling

myClassifier.fromJSON(state);

drewinglis/automator