It is amazing what we can do with smartphones these days. Document scanning and text recognition are nothing new. But being able to have such a functionality in our pockets is pretty neat. These days we can create apps that have such features very quickly thanks to the push Apple has been doing to promote Machine Learning and Artificial Intelligence on their devices.
Starting on iOS 11, we can natively scan documents with a system framework called VisionKit, and we can perform operations on images using a framework called Vision. It wasn’t until iOS 13 that we finally had the ability to recognize text on images ourselves using the Vision framework, without leveraging third party libraries. In this article we will explore how we can use the VisionKit framework to scan documents and the Vision framework to detect text as two separate tasks, so you can see how easy these two tasks are and you can learn to put them together.
If you have been using the keychain on your iOS apps you may want to start using Face ID/Touch ID to let your user access your app and their data. This is a common use case but it’s very easy to do incorrectly.
Apple introduced Touch ID all the way back in 2013, and ever since then, every iOS device has come with some sort of biometric authentication method, be it Touch ID or Face ID. This has allowed developers to implement convenient unlocking into their apps to access sensitive data without having to ask for the passcode. If your app “locks” access in any way your users are probably expecting to “unlock” with their finger or Face ID, so it is your responsibility to implement in a way that is secure and can’t be vulnered.
This article is an entirely rewritten version of an old tutorial I wrote years ago titled “Using the iOS Keychain”. Originally written in Objective-C, the old version has been archived but it is accessible here.
The Keychain is the place where you would store sensitive data. As secure as iOS currently is, the keychain is the right place to store passwords, authentication tokens, and other sensitive data. You should not store this kind of data in UserDefaults, even if iOS has made it harder to access that data for normal users in the latest versions.
There may be cases in which you need to find related words to others. With the NSLEmbedding class, you can find related strings based on the proximity of their vectors.
Using NLEmbedding
Using NLEmbedding is very straight forward. A simple task is to get an array of related words, which come as an array of (String, NLDistance) back.
The distance between words tells you how “related” they are
In the past few weeks, we have explored how we can tokenize natural language text and how to recognize the language a natural language text is written in. This week we will continue exploring more natural language APIs provided by the NaturalLanguage framework. We will learn about the NLTagger class, which allows us to to analyze natural language text to find parts of speech, lexical classes, lemma, scripts, and more. This API, introduced in iOS 12, implements machine learning to work, and just like the other NaturalLanguage classes, is very easy to use.
Continuing my trend of writing about language processing, today I want to discuss about identifying the language of a body of text. This is an interesting task we can do thanks, once again, to Apple’s investment in APIs linked to machine learning.
Today we will explore the NLLanguageRecognizer object. Introduced in iOS 12, this class can do a lot of language recognizing, from detecting the “dominant language” of a string, to all the possible languages.
Working with Natural Language is possible thanks to machine learning. Starting on iOS 12, Apple has provided many APIs just for this task. In this article we will explore how to use NLTokenizer to separate natural language text into its proper units.
Introduction to Natural Language Tokenizing
If you are not familiar with the inner workings of Natural Language processing, tokenizing simply means that we separate a string and analyze it to find its semantic units. If you are writing a program that processes text, you may be tempted to split the string using a separator. For example, if you wanted to get all the words in a natural sentence string in an array, you would write something like this:
iOS has a lot of APIs that deal with natural language detection. One such class is NSDataDetector. This class allows you to match different kinds of data in text, including dates, time, links, and more. This class, actually introduced a very long time ago (in the iOS 4.0 days!) makes it very easy to find this kind of data in strings. In this article we will explore how to use this very old class - whose documentation is Objective-C only at this time - in Swift, and how to do common tasks with it.
As iOS becomes more advanced, features that we thought belonged to the long future start becoming more common place in today’s software. One such feature is speech recognition, which allows a device to take verbal input from a user, transcribe it into text, and do something with it.
In iOS, we can do this using a framework called Speech, and an object called SFSpeechRecognizer. With this class, you can perform all kinds of speech recognition tasks.
Apple’s CryptoKit introduced this year is full of amazing features. Not only does it offer very easy to use cryptography, but it also offers an interface to a security feature that Apple introduced less than a decade ago: The Secure Enclave.
The Secure Enclave is a hardware feature for helping the system work with cryptographically secure data. In this article, we will build upon our previous CryptoKit knowledge (see the article linked above), and we will also learn what the Secure Enclave is all about.
