OffensiveAudioClassifier

Uses voice recognition to transcribe and detect offensive content in audio.

Overview

The "OffensiveAudioClassifier" library provides an innovative solution for recognizing offensive content by collecting real-time audio. It harnesses the power of machine learning and Apple's Speech framework to transcribe audio into text and classify it into three categories: "neither" (neutral), "offensive," and "hate" (hate speech).

• Audio Transcription: Transcribes audio into text with high precision, utilizing Apple's Speech framework.
• Offense Detection: Identifies offensive language and hate speech in the transcribed text.
• Machine Learning Model: Implements a robust machine learning model trained with over 35.000 expressions, using Apple's Create ML.
• Supports both SwiftUI and UIKit.
• Does not require internet connection for use.
• English language support only. (Will support more languages in the future)

Benefits

• Content Moderation: Facilitates content moderation on online platforms such as forums, social networks, and messaging apps.
• Combatting Hate Speech: Protects users against offensive and discriminatory content.
• Enhanced User Experience: Provides a more enjoyable and positive experience for your users.

Requirements

• iOS 17.0+
• macOS 14.0+
• watchOS 10.0+
• tvOS 17.0+

Installation

Swift Package Manager

1. File > Add Package Dependencies... >
2. Add https://github.com/erickrib/OffensiveAudioClassifier

OR

Update dependencies in Package.swift

dependencies: [
    .package(url: "https://github.com/erickrib/OffensiveAudioClassifier", .upToNextMajor(from: "1.0.5"))
]

Cocoapods

OffensiveAudioClassifier is available through CocoaPods. To install it, simply add the following line to your Podfile:

pod 'OffensiveAudioClassifier'

Link-pod

Usage

Next, import the library into your Swift code:

import OffensiveAudioClassifier

SwiftUI

Initialization

For SwiftUI applications, before using any methods of the library, you need to first instantiate the class. You can use the @StateObject property wrapper to manage the lifecycle of the OffensiveAudioClassifier instance:

@StateObject var offensiveClassifier = OffensiveAudioClassifier()

Features:

// You can initialize the OffensiveAudioClassifier object with an initial transcript.
@StateObject var offensiveClassifier = OffensiveAudioClassifier(initialTranscript: "example text inital")

// Initiates the transcription process to convert audio to text. Begins collecting audio input from the microphone.
offensiveClassifier.transcribe()

// Stops the transcription process.
offensiveClassifier.stopTranscribing()

// Accesses the machine learning model directly. Detects offensive content by providing a string input.
offensiveClassifier.detectOffensive(message: "exemple text")

// Accesses the text classification property that can indicate "neither," "offensive," or "hate" classifications.
offensiveClassifier.textClassification

// Property provides the real-time transcription of audio collected from the microphone.
offensiveClassifier.transcript

UIKit

Initialization

Before using any method from OffensiveAudioClassifier, instantiate the main class as a state object to manage the library's lifecycle.

import UIKit
import OffensiveAudioClassifier

class ViewController: UIViewController {
    var offensiveClassifier: OffensiveAudioClassifier?

    override func viewDidLoad() {
        super.viewDidLoad()
        offensiveClassifier = OffensiveAudioClassifier()
    }
}

The OffensiveAudioClassifier class provides two properties: transcript and textClassification. You can observe these properties to get real-time updates on the transcribed text and its classification.

To observe these properties, you can create a custom observer by conforming to the OffensiveAudioClassifierDelegate protocol:

import UIKit
import OffensiveAudioClassifier

class ViewController: UIViewController, OffensiveAudioClassifierDelegate {
    var offensiveClassifier: OffensiveAudioClassifier?

    override func viewDidLoad() {
        super.viewDidLoad()
        offensiveClassifier = OffensiveAudioClassifier()

        // Sets the delegate of the offensiveClassifier to the current object.  
        offensiveClassifier?.delegate = self
    }

    func updateTranscript(_ text: String) {
        // Handle the updated transcript
    }

    func updateOffensiveText(_ text: String) {
        // Handle the updated offensive text classification
    }
}

To start transcribing audio, call the transcribe() method

offensiveClassifier?.transcribe()

To stop transcribing, call the stopTranscribing() method:

offensiveClassifier?.stopTranscribing()

Permissions

To use the voice recognition and audio recording features, your application must request and obtain the necessary permissions from the user. Ensure that you have added the NSMicrophoneUsageDescription and NSSpeechRecognitionUsageDescription keys to your application's Info.plist file with appropriate descriptions.

Example

Key	Type	Value
Privacy - Speech Recognition Usage Description	String	This app needs access to the microphone to record and share audio in real-time.
Privacy - Microphone Usage Description	String	This app requires access to speech recognition to transcribe and analyze audio content.

Attention: To use the voice recognition and audio recording features, your application must request and obtain the necessary permissions from the user. Without these settings, the library will not function correctly.

Creators

• Erick Ribeiro
• Gabriel Eduardo
• Letícia Malagutti
• Victor Inácio