N-Gram
An n-gram is a continuous sequence of 1, 2, 3 or 4 words found in answers. It shows which formulations come back often.
It is useful to measure volume, spot frequent expressions and compare questions or groups.
Product
Harmate tracks volume, isolates noise, groups close formulations and keeps source verbatims visible so your analysis stays verifiable.
Designed for open questionnaires, response imports and analyses where volume makes manual reading unreliable.
API Demo
The text you type is sent to the Harmate API and displayed in the same volumetry widget as the dashboard: n-grams, syntactic dependencies, performant mode and noise filters. No result is stored in the database.
Stateless API processing: the content is used only for immediate computation, protected by CAP and request limits, with no database write.
How It Works
Volumetry is not just word counting. Harmate turns free-text answers into reviewable signals: repeated expressions, grammatical links, isolated noise and traceability back to source verbatims.
I like framing discussions and reformulating tensions.
An n-gram is a continuous sequence of 1, 2, 3 or 4 words found in answers. It shows which formulations come back often.
It is useful to measure volume, spot frequent expressions and compare questions or groups.
A syntactic dependency describes the grammatical link between two words: subject, verb, object, complement or coordination. It shows which word depends on which.
It is useful to preserve meaning when two answers do not use exactly the same words or word order.
Performant mode keeps expressions that carry a real signal and removes artifacts that artificially inflate volumes.
It is useful to read quickly without confusing numeric noise, overly generic words and real answer signals.
One word: useful for a first count, but often too weak without context.
Two close words: the signal becomes more precise and less noisy.
Three continuous words: useful to retrieve an almost exact formulation.
The graph links words by grammatical function. It is close to what parsing visualizations such as spaCy show: arcs above words to explain the sentence structure.
Unlike an n-gram, a dependency does not only look for neighboring words: it looks for the relationship that carries meaning.
Response Volume
By Question
See where answers are dense, weak or too dispersed.Cleaning
Noise Isolated
Set aside numeric or weak signals without losing source traceability.Review
Source Verbatims
Go back to original answers to validate an interpretation.Collected or imported answers are grouped by questionnaire, question, group and respondent to avoid partial readings.
Close expressions, typography variants and neighboring formulations are grouped to reveal real volumes.
Very short answers, numeric artifacts and contextless signals are identified so they do not pollute conclusions.
Themes, frequent expressions and representative verbatims are connected to turn raw text into a structured reading.
Each indicator remains tied to original answers so you can check a synthesis before deciding or sharing.
Identify which questions produce enough material and which ones remain too thin to support a decision.
Spot recurring formulations, their close variants and the contexts in which they appear.
Prevent an isolated number, empty answer or weak expression from carrying more weight than it should.
Keep a reliable view when the same respondent has multiple submissions or when an import contains duplicates.
Distinguish complete answers, usable verbatims and areas where you need to relaunch or rephrase.
Keep the link between synthesis, theme, expression and original answer to avoid opaque interpretations.
No. Cleaning produces a usable reading, but original answers remain available to verify signals, themes and summaries.
Typical examples include empty answers, isolated numbers without context, verbatims too short to interpret alone, or duplicates that artificially distort volume.
No. The need becomes obvious with hundreds of answers, but cleaning also helps as soon as several open questions, groups or imports need comparison.
No. It reduces noise and structures the material so human analysis starts from a more reliable, verifiable and shareable base.
Volume tracking, cleaning, grouping and source verbatims: Harmate helps you turn a mass of answers into reasoned decisions.