Tone & Emotion Analyzer

A sentiment count and six tone scores read off a fixed lexicon, with the words that caused each score shown. It is a measurement rather than a judgement, which is what makes it useful for checking whether an email reads the way you intended before you send it.

Six Tones, Counted Word by Word

Tone is the thing writers are worst at assessing in their own drafts, because they know how a sentence was meant. A count of the words that carry emotional weight is a crude instrument, but crude and inspectable beats confident and opaque when the question is whether a message will land as intended.

• Sentiment splits into three shares. Words are matched against a positive list of one hundred and seventeen entries and a negative list of one hundred and forty-four, and everything unmatched is counted as neutral, so the neutral figure is a residual rather than a third classification. The three percentages always sum to a hundred.
• The overall reading is decided by the gap between the positive and negative shares rather than by a ratio. A skew of eight percentage points or more toward positive returns positive, eight or more toward negative returns negative, and anything narrower returns neutral — a band deliberately wide enough that a single emotional word in a short text does not flip the verdict.
• Six tone categories are scored independently, each against its own word list: Formal, Friendly, Angry, Worried, Joyful and Humorous. A text can score on several at once, which is the common outcome — a formal complaint is both formal and angry, and reporting only one of those would be misleading.
• Each tone has an intensity figure derived from the number of matches per hundred words, scaled so that a small number of hits produces a meaningful reading on a short text. Intensities are capped, so a passage saturated with one kind of vocabulary cannot exceed the top of the scale.
• The matched words are listed per category, and this is the substantive part of the output. Seeing that a draft scored as angry because of three instances of the word unacceptable tells you something a percentage cannot.
• A sample button loads a passage so the layout can be seen before you paste anything of your own.

Using It as a Pre-Send Check

1. Paste the message exactly as you intend to send it, including the greeting and the sign-off. Both carry tone weight, and a draft checked without its opening line is being checked on the wrong text.
2. Read the sentiment shares first, then the tone list. If the overall reading is neutral but Angry is the highest tone, the message is more pointed in places than its average suggests — which is exactly the mismatch people get caught by.
3. Open the matched words for the tone you did not intend. Seeing the specific terms is what turns a vague sense that a draft feels off into a specific edit.
4. Change those words rather than the whole draft. Tone scores are word counts, so the score responds to vocabulary; whether the resulting message is better is still your judgement.
5. Re-run after editing, then send. The value of the check is that it happens before the send, not after the reply.

A Complaint That Reads Angrier Than Intended

Take a three-sentence email about a delayed delivery, written by someone who is mildly annoyed and wants to sound measured.

• The overall sentiment comes back negative, driven by the negative list rather than by anything extreme — delayed, unfortunately and disappointing are all entries, and three hits in a short text is a large share. This is the first useful signal: the writer's own sense of being mild does not survive contact with the arithmetic.
• The tone scores show Worried and Angry both firing, with Angry marginally higher. The matched words explain it: unacceptable and never are both in the anger list, and the writer used them to describe the company's process rather than the person, without noticing that the words do not distinguish.
• Replacing unacceptable with not what was agreed and never with has not, so far, keeps the complaint explicit while removing the two terms that drove the anger score. The sentiment reading stays negative, because the text is still a complaint and should be.
• The Formal score stays high throughout, which is worth noticing: formality and anger are independent, and a message can be perfectly polite in register while scoring high on anger vocabulary. That combination is what reads as cold rather than cross, and it is the most common failure in professional complaint emails.
• The neutral share is the largest of the three in every version, which is the normal result for transactional writing and a reminder that the residual figure is not measuring anything in particular.

What Lexicon Counting Gets Right and Wrong

Alternative	Notes
Negation is invisible	A word is counted because it appears, without regard to what precedes it. "The result was not good" contains a positive entry and scores toward positive. This is the single largest source of wrong readings, and there is no way to fix it by editing the lists — it would require parsing the sentence.
Sarcasm and irony invert	No method based on vocabulary can detect them, and no local tool will. If your writing uses irony deliberately, read the output as a description of the words rather than of the effect.
Domain vocabulary is neutral	Technical, legal and medical terms are almost entirely absent from both lists, so a clinical report scores as neutral. That is usually correct and occasionally hides a sentence that a human reader would find alarming.
Intensity is relative, not absolute	Because the score is matches per hundred words, the same three emotional words produce a much higher reading in a two-sentence message than in a thousand-word document. Comparing intensities between texts of different lengths is not meaningful; the useful comparison is the same document before and after an edit.
What it does reliably	It is consistent, inspectable and fast. The same text always produces the same reading, and every point of that reading can be traced to a specific word. As a check on whether a draft contains words you did not consciously choose, that is enough.

Where a Word Count Helps

• Before sending a message that matters — a complaint, a rejection, a negotiation — to see whether the vocabulary matches the posture you intended.
• Reviewing support replies at volume, where a consistent negative skew across many responses is a signal about tone policy rather than about any one message.
• Editing your own writing for register when you know you have a habit of over-softening or over-sharpening. The matched-words list is where the habit becomes visible.
• Not for measuring how a reader will feel. Sentiment is not emotion, and the gap between what a text contains and what a person experiences is where every method of this kind fails.
• Not for screening user-generated content at any scale. The negation problem produces false positives on precisely the sentences that most need a human to read them.
• Not for literary analysis, where irony and tension between surface and meaning are the substance rather than noise.

Questions About the Reading

It says my text is positive but it reads as negative to me. Why?

Almost certainly negation. The analyzer counts words that appear, without reading what comes before them, so a sentence like "this was not a good outcome" contributes to the positive count through the word good. The matched-words list will show you the terms responsible, and if they are sitting behind a not or a never, the reading is wrong in a way you can correct for by reading the list rather than the percentage.

How is the overall positive, negative or neutral decision made?

By the gap between the positive and negative shares. If positive leads negative by eight percentage points or more the text reads positive, if negative leads by eight or more it reads negative, and within that band it reads neutral. The band exists because a short message can swing sharply on one word, and calling a two-sentence note negative because it contains one is rarely useful.

What are the six tone categories counting?

Each has its own word list, and every category is scored independently rather than assigned as a winner. Formal, Friendly, Angry, Worried, Joyful and Humorous are measured simultaneously, which is why a text showing high Formal and high Angry is not a contradiction — that combination is what a firm, well-mannered complaint actually looks like, and knowing you produced it is more useful than being told the dominant tone.

Does any of this involve sending my text anywhere?

No. The whole analysis is word matching against lists compiled into the page, and the page contains no AI integration at all — there is no provider panel on it. Your text is never transmitted and the tool works offline, which matters here more than usual because the messages people check for tone are often the ones they would least like stored somewhere.

Last verified: October 3, 2026

Launch Tone Emotion Analyzer

All Tools · AI Tools · About · How We Test · Contact · Privacy

Maintained by Huang Bin · Last updated October 3, 2026