npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

@lanred/gibberish-detective

v1.0.4

Published

Detects if a string appears to be gibberish, using markov-chaining

Downloads

13

Readme

Gibberish Detector

This utility is used to identify if a string of text contains gibberish, using a relatively simple implementation of Machine Learning and Markov Chaining. The utility is trained with a large sample of legitimate sentences with which it builds a model to determine the probability that any two characters would be found adjacently.

Configuration

Options can be set during or after initialization:

const gibberish = require("gibberish-detective")({useCache: false});
gibberish.set("useCache", true)

Options

|Name|Type | Default | Description |--|--|--|-- |useCache | bool | true | Use cache during matrix querying |model | obj | shipped model | A learning model outputted from .train()that is used to calculate letter pairing frequencies |thresholdFn | fn(model) | avg(baseline.good.min, baseline.bad.max) | A function that outputs a number to determine the minimum value that an .assignScore() output can be before it is considered to be gibberish

Detection

The .detect()(.isGibberish() is an alias) function takes a string and returns a boolean on whether or not it detects the string as gibberish.

gibberish.detect('Luke, I am your second cousin!'); // returns false
gibberish.detect('fasdfhaiufaewroawifasdaeta'); // returns true 

You may override the training model it is testing against by passing it in as the second (and optional) argument.

Training

This package is shipped with a default model that was trained against a Sherlock Holmes novel. If necessary, a new model can be generated. As an example, if someone wishes to use this with another language, it might be prudent to generate a model based off of this language.

For best results the training text used should be decently long (at least few megabytes) with diverse and legitimate words. It is not recommended that it be trained using literature from the science fiction and fantasy genres because those selections often contain unusual word formations.

To create a new learning model, you must provide it with the aforementioned large sample of text to build the learning matrix, a sample of short lines of sentences that make sense, and a sample of lines made up of gibberish. It uses the last two data sets to form a threshold between gibberish and non-gibberish text.

let gibberish = require("gibberish-detective")();
let fs = require('fs');
let sample = fs.readFileSync('./data/good.txt', 'utf-8');
let sample_good_small = fs.readFileSync('./data/good_sm.txt', 'utf-8');
let sample_bad = fs.readFileSync('./data/bad.txt', 'utf-8');

let newModel = gibberish.train(sample, sample_good_small, sample_bad);

gibberish.set("model", newModel);

It is recommended that if re-training is necessary that this process only be done once and the resultant model be saved to a file and then thereafter that file is loaded into into the module at init-time:

At train time:

let fs = require('fs');
let newModel = gibberish.train(sample, sample_good_small, sample_bad);
fs.writeFileSync("mymodel.json", JSON.stringify(newModel))

At run-time

let myLearningModel = require('mymodel.json');
let gibberish = require("gibberish-detector")({model: myLearningModel});

Attributions

rrenaud originally created a projected similar to this in python, called Gibberish-Detector.