moderation

moderation is a profanity filter for Go.

Note: This project is not actively maintained, as I have moved on to working on a Rust version.

Goals

1. Easy to use
2. Minimum possible allocations, processing time, and binary size
3. Minimum false negatives (including text like h3110_w0r!d)
4. Minimum false positives
5. (Experimental) Provide a way to censor text
6. (Future) Other analysis types than inappropriate, profane, offensive, sexual, mean, spam (violence, contact info, etc.)
7. (Future) Basic support for languages other than English

Example

package main

import (
	"fmt"
	"github.com/finnbear/moderation"
)

func main() {
	printResult("hello world")
	printResult("$#1t")
	printResult("a$$")
	printResult("assassin")
}

func printResult(phrase string) {
	description := "is appropriate"
	if moderation.IsInappropriate(phrase) {
		description = "is NOT appropriate"
	}
	fmt.Printf("\"%s\" %s.\n", phrase, description)
}

$ go run hello_world.go
"hello world" is appropriate.
"$#1t" is NOT appropriate.
"a$$" is NOT appropriate.
"assassin" is appropriate.

Comparison

Accuracy was evaluated based on the first 100,000 items from this dataset of moderated comments.

Package	Time	Accuracy	Comment
finnbear/moderation	1.63s	92.21%	Not stable, nor actively maintained
TwinProduction/go-away	2.16s	82.18%	Many false positives from combined words like "push it"

Acknowledgements

1. Radix implementation based on https://gitlab.com/caibear/go-boggle/
2. Some profanities and test cases are based on https://github.com/TwinProduction/go-away