Smarter Keyword Analysis for PPC and SEO

How N-grams, Levenshtein Distance, and Jaccard Similarity Help SEO and PPC

Now AI may generate keywords within minutes, and even launch a paid search campaign within minutes. However, it is not sufficient to merely apply AI. To have your SEO or Google Ad campaigns work in reality, you should learn how search works and how to format your keywords.

The following are three easy ways of making sense of search data by marketers:

• N-grams: Dissecting keywords into smaller segment.
• Levenshtein Distance: Searching for spelling errors or other words.
• Jaccard Similarity: Finding the similarity between two phrases.

N-grams

This is the division of a keyword into a sequence of words.

Keyword: “private caregiver nearby”

• Unigrams (1 word): private, caregiver, nearby
• Bigrams (2 words): private caregiver, caregiver nearby
• Trigrams (3 words): private caregiver nearby

Why use N-grams?

• N-grams can be used in cases where you have 100,000 search terms and you want to identify the words or phrases that work well. And even such bad words as [free] (those who type [free] in the search engine mostly do not purchase it) can be turned into bad keywords.
• You can recognize the words that do well such as the ones that contain near and develop campaigns targeting them.
• You can identify high-performing words like “nearby” and create campaigns targeting them.

N-grams are used to transform the sloppy search data into small and useful groups, thus that you can refine your campaigns or content.

Levenshtein Distance

Levenshtein Distance is used to indicate the number of letters required to be modified in order to convert a single word into another.

• cat → cats = distance 1 (just add “s”)
• cat → dog = distance 3 (c→d, a→o, t→g)

Use in SEO and PPC:

• Find misspelled brand or competitor keywords. Example: uber → uver (distance 1)
• Combine very similar keywords to simplify ad groups.

The following is an example of grouping keywords using the Levenshtein Distance:

The meaning of all three keywords is the same. You can monitor them under a single ad group. Reporting, bidding and control of campaigns become so easy.

Jaccard Similarity

Jaccard similarity compares two phrases on the number of similar words. It does not really matter how the words are ordered.

• “new york plumber” & “plumber new york” = similarity 1 (all words same, order different)
• “new york plumber” & “NYC plumber” = similarity 0.25 (only “plumber” is common)

Use: Combine similar keywords that are just reordered. Note: It won’t understand meaning (NYC vs New York), but it works well for duplicates.

Combining Techniques for Campaigns

You can use these three techniques together for bigger campaigns:

1. Use N-grams to find high-performing words and phrases.
2. Use Levenshtein Distance to combine very similar keywords.
3. Use Jaccard Similarity to remove reordered or slightly different duplicates.

Example: Cybersecurity keywords:

• cybersecurity courses
• cybersecurity online course
• free cybersecurity courses
• online cybersecurity courses
• cybersecurity course
• google cybersecurity course

After combining, you can make four main keyword groups:

• Cybersecurity courses
• Cybersecurity courses online
• Free cybersecurity courses
• Google cybersecurity course

This keeps campaigns simple, structured, and easier to manage.

Quick Summary

Conclusion: N-grams, Levenshtein Distance and Jaccard Similarity are required to develop structured, high-performing campaigns in AI in real time, creating keywords. These methods purify your data, eliminate duplicates and enhance your campaigns to be profitable.