How to identify parts of speech and names in a text with Swift

As we learned in the developer guide on the Natural Language framework, it supports linguistic tagging. Linguistic tagging, also known as part-of-speech tagging, is the process of assigning grammatical tags to words in a text based on their role and function within a sentence.

We could use linguistic tagging for example to identify parts of speech in a text like verbs and nouns, or names like people names, place names or organization names in texts.

Let's look at how it works.

Capture HTTP(s) traffic with Proxyman

Proxyman - Your ultimate man-in-the-middle proxy to effortlessly capture, inspect, and manipulate HTTP(s) traffic on macOS, Windows, iOS, and Android devices.

Get started for free

The linguistic tagging functionality is provided by the Natural Language framework's NLTagger type. The tagger is initialized with a scheme of type NLTagScheme which specifies what kind of tags should be used for classification.

For example, we could use .tokenType as a scheme, which would classify parts of a text with the following tags: .word, .punctuation, .whitespace and .other.

Another example for a scheme is .lexicalClass, which classifies parts of a text with the tags .noun, .verb, .adjective, .adverb etc.

Let's look at the following function which extracts all verbs from a text:

func getVerbs(from text: String) -> [String] {
    let tags = tags(for: text, tagScheme: .lexicalClass)
    return tags.filter { $0.tag == .verb }.map { $0.word }
}

The function uses the following helper function which creates pairs of words and their tags based on the provided text and scheme.

func tags(for text: String, tagScheme: NLTagScheme) -> [(word: String, tag: NLTag)] {
    var taggedWords: [(String, NLTag)] = []

    let tagger = NLTagger(tagSchemes: [tagScheme])
    tagger.string = text
    tagger.enumerateTags(in: text.startIndex..<text.endIndex,
                            unit: .word,
                            scheme: tagScheme,
                            options: [.omitPunctuation, .omitWhitespace]) { tag, tokenRange in
        if let tag = tag {
            taggedWords.append((String(text[tokenRange]), tag))
        }
        return true
    }
    return taggedWords
}

We can use the same helper function to extract other words, for example names. The following code extracts place names from a given text:

func getPlaceNames(from text: String) -> [String] {
    let tags = tags(for: text, tagScheme: .nameType)
    return tags.filter { $0.tag == .placeName }.map { $0.word }
}

And that's basically it. The NLTagger makes it really easy for us to identify parts of speech in a text.

Check out the official NLTagScheme documentation to learn more about supported classification possibilities.

Capture HTTP(s) traffic with Proxyman

Proxyman - Your ultimate man-in-the-middle proxy to effortlessly capture, inspect, and manipulate HTTP(s) traffic on macOS, Windows, iOS, and Android devices.

Get started for free