All Blogs
Data Science

YAKE Keyword Analysis: A Simplified Guide for NLP Enthusiasts

April 8, 2024
min read

Keyword extraction underpins numerous NLP tasks such as summarization, information retrieval, and content analysis. YAKE (Yet Another Keyword Extraction) provides a robust solution in this domain. This unsupervised approach automates keyword extraction, identifying the most semantically significant terms within a single document.

Unlike some methods, YAKE keyword extraction does not rely on external resources like dictionaries or pre-trained models. It leverages statistical features within the text, making it adaptable to various languages, domains, and document sizes.

This makes YAKE versatile and applicable to multiple languages, domains, and text lengths without extensive training data.


 Key Features of YAKE

YAKE implements a unique unsupervised approach to keyword extraction, relying on analyzing features within the text. Here is a breakdown of its core functionalities:

1. Multi-word Phrase Handling

YAKE keyword extraction is not restricted to single words. It generates candidate keywords by considering contiguous sequences of 1, 2, and 3 words (1-grams, 2-grams, and 3-grams). This allows YAKE to capture important multi-word phrases that carry thematic meaning.

2. Contextual Adaptability

While position and frequency play a role, YAKE keyword extraction does not solely rely on them. YAKE can identify context-specific keywords by considering how often words co-occur with their surrounding terms. Words frequently appearing with a variety of surrounding words might be less meaningful, indicating they might function similarly to stop words. This allows YAKE to adapt to the specific vocabulary and phrasing within a document.

How YAKE Works

YAKE tackles keyword extraction through a multi-step process that analyzes the text without relying on external resources:

1. Text Preprocessing

YAKE begins by preparing the text for analysis. It breaks the text into individual terms, separating words based on spaces, punctuation marks, or line breaks. For instance, when YAKE processes the sentence "Natural Language Processing (NLP) is a field of Artificial Intelligence...", it breaks it down into single words such as "Natural," "Language," "Processing," and so on.  

2. Feature Extraction

YAKE goes deeper than identifying terms by extracting five key features for each word to assess its significance:

  1. Casing: YAKE takes into account whether a word is in uppercase or lowercase. For instance, "Machine" might be less indicative than "MACHINE LEARNING," suggesting the latter could be a keyword.
  2. Word Positional: YAKE assigns a higher score to words appearing earlier in the text. In the above example sentence, "Natural Language Processing" might score higher than "Artificial Intelligence" due to its position.
  3. Word Frequency: Words appearing more frequently receive a higher score. "The" would likely score low due to its frequency, while "Intelligence" might score higher.
  4. Word Relatedness to Context: YAKE analyzes how many different words surround a candidate keyword. Consider "learning": If it co-occurs with many unique words (e.g., "machine," "deep," "supervised"), it might be less meaningful than "machine learning," which appears together frequently. 
  5. Word DifSentence: This assesses how often a word appears in different sentences. Words found in multiple sentences (e.g., "intelligence") are considered more relevant than those restricted to one sentence (e.g., "a").

3. Scoring and Candidate Generation

YAKE keyword extraction combines these features into a single score for each term and each candidate phrase (1-grams, 2-grams, and 3-grams). Lower scores indicate potentially more relevant keywords.  

4. Data Deduplication and Ranking

YAKE removes duplicates and ranks the remaining keywords based on their scores. It provides a final list of the most relevant keywords and key phrases, summarizing the document's content.

YAKE analyzes the text's structure, word usage, and co-occurrence patterns to identify the most informative keywords representing the document's content.

YAKE vs. Other Keyword Extraction Methods

While YAKE offers a powerful approach, it is valuable to compare it with other keyword extraction techniques:

Traditional Methods 

Advantages: Traditional methods like TF-IDF and RAKE are often simpler and computationally less expensive. TF-IDF, for instance, focuses on word frequency within a document and across a corpus, offering a clear interpretable score.

Limitations: They can be influenced heavily by document length and may struggle with context or identifying multi-word phrases. Relying on pre-built dictionaries or corpora might limit their adaptability to new domains or languages.

Advantages of YAKE Keyword Extraction

  • Unsupervised Approach: YAKE's unsupervised nature eliminates the need for training data or external resources.
  • Focus on Text Features: Analyzing features within the text allows YAKE to capture context-specific keywords and potentially identify emerging terminology not yet established in external resources.

Limitations of YAKE Keyword Extraction

  • Context Ambiguity: YAKE's reliance on local text statistics can make it susceptible to ambiguity and context dependency. Complexities in natural language can still pose challenges in accurately identifying the most relevant keywords, particularly in very short or very long documents.
  • Multi-word Phrase Identification: While YAKE considers n-grams, identifying highly specialized multi-word phrases might be less efficient compared to methods that leverage domain-specific knowledge.

YAKE keyword extraction balances efficiency and adaptability, making it a strong choice for many NLP tasks. However, traditional methods like TF-IDF might be preferred for situations requiring maximum interpretability or dealing with highly specialized domains.

YAKE Implementation and Examples

YAKE's versatility shines in various NLP tasks. Here are some examples:

  • Automatic Document Summarization: YAKE helps create concise summaries that capture the essence of a document by identifying key terms. In a lengthy research paper, YAKE can extract keywords like "gene editing", "ethical implications", and "potential benefits", aiding in summarizing the paper's core themes.
  • Information Retrieval: YAKE can improve search engine functionality. When a user enters a query, YAKE can extract relevant keywords from the indexed documents, allowing for more accurate retrieval of documents matching the user's intent.

Implementing YAKE with Python Libraries

Fortunately, leveraging YAKE in Python is straightforward. Here is a glimpse of how it might look:

from yake import KeywordExtractor

# Text to be analyzed
text = "This is a sample document about natural language processing (NLP)."
# Define the extractor
extractor = KeywordExtractor()
# Extract keywords with a maximum of 2 words per phrase
keywords = extractor.extract_keywords(text, top=5, n=2)
# Print the extracted keywords

This code snippet demonstrates how to extract the top 5 keywords, considering phrases up to 2 words long. The output would likely include terms like "natural language processing" and "keyword extraction", showcasing YAKE's ability to identify key multi-word phrases.

Understanding YAKE Output

Here is a Python code snippet demonstrating YAKE's output format and relevance scores, along with a different document for analysis:


from yake import KeywordExtractor

# Sample document about sentiment analysis

text = """Sentiment analysis is a branch of Natural Language Processing (NLP) that focuses on

        identifying the sentiment of the opinion expressed in a piece of text. It

        determines whether the sentiment expressed is positive, negative, or neutral. 

        Sentiment analysis is widely used in various applications, including social media

        monitoring, market research, and customer service."""

# Define the extractor

extractor = KeywordExtractor()

# Extract keywords with a maximum of 3 words per phrase

keywords = extractor.extract_keywords(text, top=10, n=3)

# Print the extracted keywords and their relevance scores

for a phrase, score in keywords:

  print(f"Keyword: {phrase}, Score: {score}")


This code analyzes a document about sentiment analysis. The `extract_keywords` function extracts the top 10 keywords/phrases (up to 3 words long) and their corresponding scores. The output will be similar to:


Keyword: sentiment analysis, Score: 0.024223234377464853

Keyword: Natural Language Processing (NLP), Score: 0.0821917808219178

Keyword: sentiment expressed, Score: 0.0851063829787234

Keyword: positive, negative, or neutral, Score: 0.09428089887640449

Keyword: social media monitoring, Score: 0.10345541677508551

Keyword: market research, Score: 0.10857142857142857

Keyword: customer service, Score: 0.11368744036777164

Keyword: opinion expressed in a piece of text, Score: 0.12486486486486486


As you can see, YAKE outputs each keyword phrase along with a numerical score. Lower scores indicate higher relevance based on YAKE's assessment within the context of the document.

Selecting Effective Keywords from YAKE's Output

Refine your chosen keywords based on these principles:

  • Relevance: Prioritize keywords with lower YAKE scores, indicating higher relevance within the document's context.
  • Thematic Coverage: Aim for diverse keywords that capture the document's core themes and various aspects.
  • Task Alignment: Consider your specific application (e.g., summarization, information retrieval). Choose keywords that best support your intended use case.

Use Cases and Applications

The versatility of YAKE Keyword Extraction shines in various NLP tasks. Here are some compelling use cases:

  1. Information Retrieval: Imagine a customer service representative searching a knowledge base for solutions. YAKE can extract relevant keywords from customer queries and internal documents, enabling a search engine to retrieve the most accurate and helpful information. 
  2. Automatic Summarization: News organizations automatically generate summaries of lengthy articles with YAKE. It helps create concise summaries that capture the essence of the news story for readers on the go by identifying key terms.
  3. Content Analysis: Market research firms can utilize YAKE to analyze social media conversations about a new product launch. Extracted keywords can reveal public sentiment (positive, negative, or neutral) and identify emerging themes within customer discussions.


YAKE leverages text features for unsupervised keyword extraction, pinpointing informative single words and multi-word expressions. YAKE's unsupervised nature makes it adaptable to various domains and text lengths, offering a powerful tool for extracting vital information from textual data and enriching numerous NLP applications.

MarkovML provides a user-friendly platform to integrate YAKE into your NLP workflows. Explore our comprehensive toolkit and discover how YAKE keyword extraction can help you excel in your next project. Experiment with different parameter settings and explore their effectiveness in various NLP tasks.

From Data To GenAI Faster.

Easily Integrate GenAI into Your Enterprise.
Book a Demo

A data science and AI thought-leader

Create, Discover, and Collaborate on ML

Expand your network, attend insightful events

Join Our Community