/TextClassification

Text classification approaches in Swift 4.2. With/Without CoreML

Primary LanguageC++MIT LicenseMIT

TextClassification

Text classification approaches in Swift 5.0. With/Without CoreML

This repo has sample codes for tech talk "Classifying a text to iOS without CoreML: how and why?" at https://eatdog.com.ua on March 21, 2019 and on 15th CocoaHeads Kyiv https://cocoaheads.org.ua/cocoaheadskyiv/15 on July 28th, 2019.

Getting started

  1. Ensure you have carthage installed:
brew install carthage
  1. Install dependencies:
carthage bootstrap

  1. Run unit tests for TextClassificationMacOS target. Failing tests is expected: this wa y they display actual accuracy for classification method as an output.

  2. To face MessageFilteringExtension RAM problem, use CoreMLClassifier for message filtering in MessageFilterExtension.swift. You will have to run this extension on real iPhone, and receive a real SMS from unknown sender to trigger the extension. Debugger works more or less fine. Changing text classifier onto MemoryMappedNaiveBayesClassifier demonstrates fitting into 6Mb memory limit.

  3. If you wan't just to see text classification, run the MessageFilteringApp on either device or simulator.

Licence

This project has MIT licence. It uses Google Flatbuffer library as a dependency: https://github.com/google/flatbuffers/blob/master/LICENSE.txt