A configurable gibberish scorer.

npm install gibberish-scorer

import { GibberishScorer } from 'gibberish-scorer'
...

const gibberishScorer = new GibberishScorer(useBigram?: boolean, options?: {
    initialTrainingText?: string
    goodSamples?: string[]
    badSamples?: string[]
    additionalCharsToInclude?: string
})

Instantiates new GibberishScorer object. Constructor takes optional arguments useBigram (defaults to true: prefer bigram over trigram) and options. All fields in options default to hard-coded default values: initialTrainingText defaults to stringified J.K. Rowling's Harry Potter and the Sorcerer's Stone, goodSamples defaults to an array of a few random English sentences, badSamples defaults to a few gibberish strings, and additionalCharsToInclude defaults to the empty string. Note that additionalCharsToInclude will append chars to a default of all lower case alphabetic characters and the space character (unicodes 32 and 97-122).

GibberishScorer.train: (text: string) => void

gibberishScorer.train(my_own_training_text) // Additional training for gibberishScorer if desired

Trains the GibberishScorer with any training string passed to it. This will adjust the n-gram probabilities on top of the initial training and any prior training.

GibberishScorer.recalibrateCutoffs: (goodSamples: string[], badSamples: string[]) => void

gibberishScorer.recalibrateCutoffs(good_sample_texts, bad_sample_texts) // Recalculate predicted score cutoffs based on provided samples

Re-calculates the estimated cutoff scores.

GibberishScorer.isGibberish: (text: string, strictness?: CutoffScoreStrictness) => boolean

gibberishScorer.isGibberish('The quick fox jumps over the lazy dog') // false
gibberishScorer.isGibberish('Tom Brady') // false
gibberishScorer.isGibberish('oqbwifsiehf osdfbw sjkdoo thehwei') // true
gibberishScorer.isGibberish('This sentence is half gibberish lwpqgtyukcvi', 'Avg') // false
gibberishScorer.isGibberish('This sentence is half gibberish lwpqgtyukcvi', 'Strict') // true

Returns whether input text string is gibberish, according to trained cutoff predictions and desired strictness. strictness field must be 'Loose' | 'Avg' | 'Strict' and defaults to Avg.

GibberishScorer.getScoreInfo: (text: string) => { cutoffs: CutoffScore; score: number }

gibberishScorer.getScoreInfo('This sentence is half gibberish lwpqgtyukcvi')
// {
//  cutoffs: {
//      loose: 0.021219297921045675,
//      avg: 0.028564356546162385,
//      strict: 0.035909415171279095
//  },
//  score: 0.030018117586724725
// }

Returns current cutoffs of NGramMatrix and calculated score for inspection.

interface GibberishScorerInterface {
    NGramMatrix: NGramMatrix
    train: (text: string) => void
    recalibrateCutoffs: (goodSamples: string[], badSamples: string[]) => void
    isGibberish: (text: string, strictness?: CutoffScoreStrictness) => boolean
    getScoreInfo: (text: string) => { cutoffs: CutoffScore; score: number }
}

class GibberishScorer implements GibberishScorerInterface {}

interface NGramMatrix {
    cutoffScores: CutoffScore
    train: (text: string) => void
    getScore: (text: string) => number
    recalibrateCutoffScores: (goodSamples?: string[], badSamples?: string[]) => void
    isGibberish: (text: string, strictness?: CutoffScoreStrictness) => boolean
}

interface NGramMatrixOptions {
    initialTrainingText?: string
    goodSamples?: string[]
    badSamples?: string[]
    additionalCharsToInclude?: string
}

enum CutoffScoreStrictness {
    Strict = 'Strict',
    Avg = 'Avg',
    Loose = 'Loose',
}

MIT