Text N-gram frequency analysis & related word co-occurrence word extraction tool | ZeroTools

Count sequences of 1 to 4 Unicode code points or words split by TinySegmenter. Find words within 2, 3, 5, or 8 tokens of a target, and export frequencies and proximity rates as CSV.

Loading tool interface...

When to use this tool

Back to the tool ↑

Designed to audit redundant wording and repetitive expressions by analyzing recurring character or word frequencies. It inspects top N-grams (1–4) and exportable co-occurrences.

Try it first

Enter 'abab', select character unit and 2-gram with stopword filtering off, and view 'ab' (2 counts) and 'ba' (1 count) in the frequency tab or exported CSV.

Before you start

Token boundaries depend on the splitting rules, and counts do not establish meaning or causation. Character mode can split some emoji; co-occurrence uses a count-based score that can exceed 100%, rather than standard Jaccard similarity.

Local Browser Processing and Network Communication Notice

This tool's processing operates locally within your browser. Standard site assets, analytics, and advertisements may communicate separately. There are no guarantees.

ZeroTools: Browser Processing & Privacy

ZeroTools focuses on tools that process input on your device. Check each tool’s scope and limitations before use.

Processing and privacy policy
Article Index
Ch. 1Analysis scopeCh. 2Character N-gramsCh. 3Word boundaries and stop wordsCh. 4Frequency and CSV outputCh. 5Co-occurrence and proximity rateCh. 6Interpreting results
Chapter 1

Analysis scope

The tool analyzes up to 200,000 Unicode code points from one text. It runs in the browser and recalculates when the text or settings change. Large texts can be split for separate analysis.

Chapter 2

Character N-grams

Character mode counts overlapping windows of 1 to 4 Unicode code points. For example, 😀a has two 1-grams. Combined emojis and accented letters made with combining marks can span several code points.

When stop words are enabled, the specified strings are removed before counting. This can join characters that were separated in the original text.

Chapter 3

Word boundaries and stop words

Word mode uses TinySegmenter to split the text, trims each token, and excludes empty tokens and exact, case-sensitive matches from the comma-separated stop-word list.

It then joins each consecutive group of N tokens with spaces. Proper names and specialist terms may split in unexpected places.

Chapter 4

Frequency and CSV output

The percentage is each sequence count divided by the total number of windows. The screen and CSV show up to 50 N-grams or 30 neighbors for the selected tab.

The CSV includes a UTF-8 BOM and uses the same proximity rate as the screen.

Chapter 5

Co-occurrence and proximity rate

Each exact target occurrence has a window of 2, 3, 5, or 8 tokens in each direction. A neighbor counts once per target window and the target word itself is excluded.

The proximity rate is the number of target occurrences with that neighbor divided by the total number of target occurrences, as a percentage.

With cat dog dog, dog appears near one of one cat occurrences (100%) and appears twice overall.

Chapter 6

Interpreting results

The window follows token positions, including positions across sentence boundaries. Compare analyses with the same text, window size and stop-word list.

Check the source context to interpret repeated sequences or related words.

Frequently Asked Questions (FAQ)

A.
Character mode counts Unicode code points. A single-code-point emoji stays intact. Combined emojis and letters with separate combining marks can contain multiple code points.
A.
It uses TinySegmenter for word boundaries, which does not parse parts of speech, base forms, or dependencies. Proper nouns and new words may be split unexpectedly.
A.
Input is limited to 200,000 Unicode code points. The browser processes the text on the main thread. Split larger texts into smaller samples.
A.
The proximity rate is the percentage of target occurrences with the neighbor in the selected window. In cat dog dog, dog is found near one of one cat occurrences, so the proximity rate is 100% and the overall dog count is two.
A.
It shows the top 50 N-grams and top 30 co-occurring words, which can be exported to a BOM-equipped CSV. It does not calculate standard Jaccard coefficients, TF-IDF, search rankings, or SEO effects.

Related Utilities

Garbled character repair/analysis & Excel CSV character code mutual conversion tool | ZeroTools

Useful for decoding the original file from your selected character encoding before passing the readable text to the N-gram analyzer. Note that there is no guarantee of recovering already garbled characters.

Bulk extraction of specific characters and information (email/URL/phone number, etc.) using regular expressions | ZeroTools

Extract known pattern matches like URLs or emails from a small trusted text to narrow down the target before checking frequency distributions. Note that this does not guarantee complete safety.

Real-time character count/line count counter (manuscript paper equivalent) | ZeroTools

Aggregate character, line, and word counts from another perspective before running intensive N-gram analysis.