This is a naive bayes spam classifier, trainer, and API server all rolled into one. It is not sophisticated but it works well for blocking basic email, comment or other text based spam.
const Classifier = require('spam-classifier').Classifier
const classifier = new Classifier()
classifier.save('path/to/model.json') // creates when not existing
classifier.load('path/to/model.json')
const spammyText = `Subject: hey there\nYou won the lottery!`
const hammyText = `Subject: check out my npm module\nIt is published live!`
classifier.trainSpam(spammyText)
classifier.trainHam(hammyText)
const result = classifier.predict('Am I a spam bot?') // true or false
classifier.save('path/to/model.json')# must be a path to a sqlite database
node cli.js train data/model.json --db=~/spam.sqlite --table=emails
node cli.js predict data/model.json file1.eml file2.eml file3.eml
$ MODEL=data/model.json PORT=3000 node server.js
Loaded classifier from data/model.json
spam server is listening on port 3000
Request example:
POST http://localhost:3000/api/spam-checks
{
to: '[email protected]',
from: '[email protected]',
subject: 'hey',
text: 'yo'
}
response example:
{
spam: false,
spamminess: 0.193776
}
MIT
See LICENSE file in this repository.