README

PHPAntiSpam是一个库，用于识别文档/消息/文本是否为垃圾邮件。该库使用统计分析。

4个步骤的解释

• 创建分词器
• 从历史消息中创建语料库（包含词素）
• 选择用于分类的方法
• 分类消息

实现的方法

• Paul Graham方法
• Brian Burton方法
• Robinson几何平均测试方法
• Fisher-Robinson逆卡方检验方法

安装

composer require bgruszka/phpantispam "^0.2"

示例

<?php

// First add autoloader and all necessary classes
require_once 'vendor/autoload.php';

use PHPAntiSpam\Corpus\ArrayCorpus;
use PHPAntiSpam\Classifier;
use PHPAntiSpam\Tokenizer\WhitespaceTokenizer;

// Let's decleare our example training set
$messages = [
    ['category' => 'spam', 'content' => 'this is spam'],
    ['category' => 'nospam', 'content' => 'this is'],
];

// As tokenizer we can use the simplest one - WhitespaceTokenizer (but of course you can also use RegexpTokenizer
// or create new one)
$tokenizer = new WhitespaceTokenizer();

// Let's define our corpus - collection of text documents
$corpus = new ArrayCorpus($messages, $tokenizer);

// For classifying text we can use different methods

// ------------------------------------------------------------------------------------
// Graham method

$classifier = new Classifier($corpus);
$classifier->setMethod(new \PHPAntiSpam\Method\GrahamMethod($corpus));

$spamProbability = $classifier->isSpam('This is spam');

echo 'With Graham method:' . PHP_EOL;
echo sprintf('Spam probability: %s', $spamProbability) . PHP_EOL;
echo sprintf('Is spam: %s', $spamProbability < 0.9 ? 'NO' : 'YES') . PHP_EOL . PHP_EOL;

// ------------------------------------------------------------------------------------
// Burton method

$classifier = new Classifier($corpus);
$classifier->setMethod(new \PHPAntiSpam\Method\BurtonMethod($corpus));

$spamProbability = $classifier->isSpam('This is spam');

echo 'With Burton method:' . PHP_EOL;
echo sprintf('Spam probability: %s', $spamProbability) . PHP_EOL;
echo sprintf('Is spam: %s', $spamProbability < 0.9 ? 'NO' : 'YES') . PHP_EOL . PHP_EOL;

// ------------------------------------------------------------------------------------
// Robinson Geometric Mean Test Method

$classifier = new Classifier($corpus);
$classifier->setMethod(new \PHPAntiSpam\Method\RobinsonGeometricMeanTestMethod($corpus));

$spamProbability = $classifier->isSpam('This is spam');

echo 'With Robinson Geometric Mean Test method:' . PHP_EOL;
echo sprintf(
    'Spam probability: [spamminess: %s; hamminess: %s; combined: %s]', 
    $spamProbability['spamminess'], 
    $spamProbability['hamminess'], 
    $spamProbability['combined']
) . PHP_EOL;
echo sprintf('Is spam: %s', $spamProbability['combined'] <= 0.55 ? 'NO' : 'YES') . PHP_EOL . PHP_EOL;

// ------------------------------------------------------------------------------------
// Fisher-Robinson Inverse Chi Square Method

$classifier = new Classifier($corpus);
$classifier->setMethod(new \PHPAntiSpam\Method\FisherRobinsonInverseChiSquareMethod($corpus));

$spamProbability = $classifier->isSpam('This is spam');

echo 'With Fisher-Robinson Inverse Chi Square method:' . PHP_EOL;
echo sprintf(
    'Spam probability: [spamminess: %s; hamminess: %s; combined: %s]', 
    $spamProbability['spamminess'], 
    $spamProbability['hamminess'], 
    $spamProbability['combined']
) . PHP_EOL;
echo sprintf('Is spam: %s', $spamProbability['combined'] <= 0.55 ? 'NO' : 'YES') . PHP_EOL;