Product

Response VolumeData CleaningOpen Responses

Turn hundreds of free-text answers into data you can actually use.

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

Type An Answer And See Real Volumetry

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.

WaitingNot Stored In Database

Expression Volumetry

Sort By:
Analysis Mode:
Expression Mode:
Size:
Minimum
Maximum
Hide Question Terms

Removes words already present in the question when the backend identifies them.

Hide Expressions With Numbers

Removes grades, percentages and other numeric artifacts.

How It Works

Understand What The Widget Measures

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.

Example Answer

I like framing discussions and reformulating tensions.

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.

Example“framing discussions” is a 2-gram.

It is useful to measure volume, spot frequent expressions and compare questions or groups.

Syntactic Dependency

A syntactic dependency describes the grammatical link between two words: subject, verb, object, complement or coordination. It shows which word depends on which.

Example“reformulating → tensions” indicates that tensions is the object of reformulating.

It is useful to preserve meaning when two answers do not use exactly the same words or word order.

Performant Mode

Performant mode keeps expressions that carry a real signal and removes artifacts that artificially inflate volumes.

Example“12/20” can be hidden, while “framing discussions” remains usable.

It is useful to read quickly without confusing numeric noise, overly generic words and real answer signals.

N-Gram Example

Ilikeframingthediscussionsandreformulatingtensions
1-Gramframing

One word: useful for a first count, but often too weak without context.

2-Gramframing discussions

Two close words: the signal becomes more precise and less noisy.

3-Gramlike framing discussions

Three continuous words: useful to retrieve an almost exact formulation.

Syntactic Dependency Diagram

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.

Syntactic Dependency DiagramThe 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.subjectcomplementobjectcoordinationobjectIsubjectlikepivot verbframingactiondiscussionsobjectreformulatinglinked actiontensionsobject

Unlike an n-gram, a dependency does not only look for neighboring words: it looks for the relationship that carries meaning.

Widget Glossary

Token
A unit split from text: often a word, sometimes punctuation or a contracted form.
Lemma
The base form of a word. For example, organize, organized and organizes can be grouped.
Occurrence
Number of times an expression appears in the analyzed answers.
Volumetry
Quantitative reading of available material: how many answers, how many expressions, where signal is dense or weak.
Noise
Element that can distort reading: isolated number, empty answer, overly generic word or contextless expression.
Question Terms
Words already present in the prompt. Hiding them prevents counting the question itself as an answer signal.
Numeric Filter
Filter that sets aside grades, percentages and isolated numbers when they do not carry usable meaning.
Source Verbatim
Full original answer used to verify that a signal or summary is not interpreted out of context.

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.

How Harmate Makes Response Volume Usable

Centralize Responses

Collected or imported answers are grouped by questionnaire, question, group and respondent to avoid partial readings.

Normalize Wording

Close expressions, typography variants and neighboring formulations are grouped to reveal real volumes.

Isolate Noise

Very short answers, numeric artifacts and contextless signals are identified so they do not pollute conclusions.

Group Signals

Themes, frequent expressions and representative verbatims are connected to turn raw text into a structured reading.

Verify Sources

Each indicator remains tied to original answers so you can check a synthesis before deciding or sharing.

What The Results Page Helps You Control

Volumes By Question

Identify which questions produce enough material and which ones remain too thin to support a decision.

Frequent Expressions

Spot recurring formulations, their close variants and the contexts in which they appear.

Noise And Artifacts

Prevent an isolated number, empty answer or weak expression from carrying more weight than it should.

Unique Respondents

Keep a reliable view when the same respondent has multiple submissions or when an import contains duplicates.

Response Quality

Distinguish complete answers, usable verbatims and areas where you need to relaunch or rephrase.

Traceability

Keep the link between synthesis, theme, expression and original answer to avoid opaque interpretations.

Common Questions About Response Cleaning

Does Harmate modify original answers?

No. Cleaning produces a usable reading, but original answers remain available to verify signals, themes and summaries.

What counts as noise?

Typical examples include empty answers, isolated numbers without context, verbatims too short to interpret alone, or duplicates that artificially distort volume.

Is it only useful for very large questionnaires?

No. The need becomes obvious with hundreds of answers, but cleaning also helps as soon as several open questions, groups or imports need comparison.

Does cleaning replace human analysis?

No. It reduces noise and structures the material so human analysis starts from a more reliable, verifiable and shareable base.

Analyze Open Responses Without Reading Everything Manually

Volume tracking, cleaning, grouping and source verbatims: Harmate helps you turn a mass of answers into reasoned decisions.