Headline analyzer
What are Emotional Words in a Headline?
Emotional words are headline tokens whose LabMT hedonometer valence falls below 4.0 (negative end) or above 6.0 (positive end). Seeklyn targets 10–15% of headline tokens as emotional; the bucket scores 100 in that band and decays at 4 points per percentage point outside it.
Seeklyn tags emotional words with a distinct chip in the Word Balance token row and shows the emotional percentage alongside its target range — visible at /dashboard/headline.
What makes a word "emotional" in Seeklyn?
A word is emotional when its LabMT hedonometer valence score falls outside the neutral band of 4.0–6.0. Words rated below 4.0 carry negative emotional weight (e.g. "death", "fail"); words rated above 6.0 carry positive emotional weight (e.g. "love", "win"). The LabMT dataset, developed at the University of Vermont, assigns human-rated happiness scores to ~10,000 English words and is widely used in computational linguistics research.
Words in the 4.0–6.0 neutral band are not counted as emotional even if they carry some feeling colloquially. The 4.0–6.0 threshold matches the hedonometer's published "neutral lens" exclusion band, keeping only words with a clear affective signal.
What is the target emotional-word percentage?
10–15% of headline tokens should be emotional. Below 10% the headline reads as flat and fails to engage the reader's feelings; above 15% it may feel overwrought or manipulative. Seeklyn scores the emotional bucket at 100 when inside the 10–15% band and penalises drift at 4 points per percentage point outside it, floored at 0.
For a 7-word headline, 10–15% works out to roughly 1 emotional word. For a 10-word headline it is 1–2 emotional words. The percentage-based target scales naturally with headline length, unlike the power-word count target which is a fixed absolute number (1–3).
How do emotional words differ from power words?
Emotional words are defined by LabMT valence; power words by Seeklyn's curated persuasion lexicon. A word can qualify as both — "amazing" has a high valence and is persuasive — but Seeklyn counts it only as power (priority: power > emotional > common > uncommon). The emotional bucket shows only words that are strongly valenced and not power words.
| Property | Emotional words | Power words |
|---|---|---|
| Defined by | LabMT valence <4.0 or >6.0 | Seeklyn curated persuasion lexicon |
| Measurement | Percentage of tokens (10–15%) | Absolute count (1–3) |
| Purpose | Emotional engagement | Persuasion / value signal |
| Overlap resolution | Counted as power if also power | N/A |
Why does emotional range matter for click-through?
Headlines with at least 1 clearly positive or negative word outperform neutral phrasing in click-through tests — emotional language activates the reader's response before they decide to click. The 10–15% ceiling prevents emotional overload, where a headline packed with valence words reads as sensational rather than credible.
Frequently asked questions
Does the LabMT valence threshold change between positive and negative words?
No — the same threshold applies in both directions. Any word with a LabMT score below 4.0 is negative-emotional and any word above 6.0 is positive-emotional. The neutral exclusion band is symmetric around 5.0.
Is emotional-word percentage the same as sentiment score?
No. Emotional-word percentage measures what share of tokens is emotionally charged (regardless of direction). Sentiment analysis uses the AFINN lexicon to compute a net directional polarity (positive, negative, or neutral) and a 0–100 strength score. The 2 signals complement each other.
Can a headline with no emotional words still score well overall?
Yes, but the emotional bucket scores 0, which pulls the word-balance score down. Word balance carries 34% of the CTR score, so a missing emotional word typically costs several points on the overall headline score. Adding 1 emotional word can flip the bucket from 0 to a positive score.