Documer

Bayes algorithm implementation in PHP for auto document classification.

Concept

every document has key words e.g. Margaret Thatcher

every document has a label e.g. Politics

Suppose, that in every document there are key words all starting with an uppercase letter. We store these words in our DB end every time we need to guess a document against a particular label, we use Bayes algorithm.

Let's clear that out:

Training:

First, we tokenize the document and keep only our key words (All words starting with an uppercase letter) in an array. We store that array in our DB.

Guessing:

This is very simple. Again, we parse the document we want to be classified and create an array with the key words. Here is the pseudo code:

for every label in DB
	for every key word in document
		P(label/word) = P(word/label)P(label) /	( P(word/label)P(label) + (1 - P(word/label))(1 - P(label)) )

Usage

Install through composer

"require": {
    "kbariotis/documer": "dev-master"
  },

Instantiate

Pass a Storage Adapter object to the Documer Constructor.

$documer = new Documer\Documer(new \Documer\Storage\Memory());

Train

$documer->train('politics', 'This is text about Politics and more');
$documer->train('philosophy', 'Socrates is an ancent Greek philosopher');
$documer->train('athletic', 'Have no idea about athletics. Sorry.');
$documer->train('athletic', 'Not a clue.');
$documer->train('athletic', 'It is just not my thing.');

Guess

$scores = $documer->guess('What do we know about Socrates?');

$scores will hold an array with all labels of your system and the posibbility which the document will belong to each label.

Storage Adapters Implement Documer\Storage\Adapter to create your own Storage Adapter.