Yake.NET

A pure C# implementation of YAKE (Yet Another Keyword Extractor) — an unsupervised, lightweight, and language-agnostic keyword extraction algorithm for single documents.

Lower score = more relevant keyword.

---

Features

• ✅ No external dependencies — 100% pure C#
• ✅ Unsupervised — no training data or models needed
• ✅ Language-agnostic — works with any language when paired with custom stopwords
• ✅ N-gram support — extracts single words and multi-word keyphrases
• ✅ Built-in English stopwords
• ✅ Deduplication of near-duplicate candidates
• ✅ Fully configurable via YakeOptions
• ✅ .NET 10

---

Installation

dotnet add package Yake.NET

---

Quick Start

using Yake.NET;

var extractor = new YakeExtractor();
var keywords = extractor.Extract("Your document text goes here...");

foreach (var kw in keywords)
    Console.WriteLine(kw); // e.g. "keyword extraction (score: 0.001234)"

---

Configuration

var extractor = new YakeExtractor(new YakeOptions
{
    MaxNGramSize          = 3,    // include up to trigrams
    TopN                  = 10,   // return top 10 keywords
    DeduplicationThreshold = 0.9, // similarity threshold for dedup (0–1)
    WindowSize            = 2,    // co-occurrence window
    Stopwords             = null  // null = use built-in English list
});

Custom Stopwords

// Extend the built-in English stopwords
var customStopwords = StopwordsEn.Words.Concat(["custom", "words"]);

var extractor = new YakeExtractor(new YakeOptions
{
    Stopwords = customStopwords
});

---

How YAKE Works

YAKE scores each candidate keyword using 5 statistical features:

Feature	Description
T_Casing	Ratio of capitalised (non-sentence-start) occurrences
T_Position	Inverse log of the sentence where the word first appears
T_Frequency	Normalised term frequency
T_Relatedness	Diversity of co-occurring words relative to frequency
T_DifferentSentence	Fraction of sentences containing the word

The final score formula for n-grams:

score(candidate) = ∏(word_scores) / (n × (n + Σ word_scores))

---

API Reference

YakeExtractor

Member	Description
YakeExtractor()	Creates extractor with default English options
YakeExtractor(YakeOptions)	Creates extractor with custom options
IReadOnlyList<KeywordResult> Extract(string text)	Extracts keywords from text

KeywordResult

Property	Type	Description
Keyword	string	The extracted keyword or keyphrase
Score	double	YAKE score (lower = more relevant)

YakeOptions

Property	Default	Description
MaxNGramSize	3	Maximum n-gram length
TopN	10	Number of keywords to return
DeduplicationThreshold	0.9	Similarity cutoff for deduplication
WindowSize	2	Co-occurrence context window size
Stopwords	null	Custom stopwords (null = English built-ins)

---

References

• Campos, R., Mangaravite, V., Pasquali, A., Jorge, A., Nunes, C., & Jatowt, A. (2020). YAKE! Keyword extraction from single documents using multiple local features. Information Sciences, 509, 257–289.

---

License

MIT