Headline analyzer
What is Headline Sentiment Analysis?
Headline sentiment analysis sums AFINN lexicon valences across headline tokens to produce a polarity label (positive, negative, or neutral) and a 0–100 strength score. Strength saturates at a net AFINN magnitude of 9.0 — approximately 2 strongly polarised words — and the strength score feeds directly into the CTR sub-score with a 13% weight.
Seeklyn displays the polarity label, strength score, and the individual positive and negative words in the Sentiment card — visible at /dashboard/headline.
What is the AFINN lexicon?
AFINN is a manually curated English lexicon that assigns a signed integer valence from −5 (very negative) to +5 (very positive) to approximately 3,500 words and short phrases. Developed by Finn Årup Nielsen, it is widely used in natural-language processing because its signed integer format allows straightforward summation over a text. Seeklyn uses AFINN for headline sentiment because it gives a signed net score that directly reflects whether the headline leans positive, negative, or neutral — and it identifies the exact words driving that polarity.
Seeklyn uses the LabMT hedonometer separately for the emotional-word bucket in word balance. AFINN and LabMT are complementary: AFINN provides signed directional scores suitable for summation; LabMT provides unsigned happiness ratings suitable for thresholding. Headlines can have a neutral AFINN net score but still contain emotional words that score above the 6.0 or below the 4.0 LabMT threshold.
How is the polarity label determined?
3 polarity labels are possible. Seeklyn sums the AFINN valence of each token in the headline. A positive net sum → positive; a negative net sum → negative; a net sum of exactly 0 → neutral. Neutral can mean no AFINN words are present or that positive and negative words cancel each other out. The polarity label and the lists of individual positive and negative words are both returned in the analysis response, so writers can see which specific words are driving the signal.
How is sentiment strength calculated?
Sentiment strength is a 0–100 score computed from the absolute net AFINN magnitude. Strength = round(100 × min(1.0, |net| / 9.0)). The cap of 9.0 means that 2 words each scoring ±4–5 on AFINN are sufficient to reach full strength (100). A headline with no AFINN words scores 0; a headline with 1 moderately positive word (e.g. "good", AFINN +3) scores about 33.
The strength score becomes the sentiment sub-score and feeds directly into the CTR score at a 13% weight. Higher strength — in either direction — is rewarded because emotionally charged headlines (positive or negative) consistently outperform neutral ones on click-through. A negative sentiment can be intentional ("Why Your Headline Strategy Is Failing") and earns the same strength credit as a positive one.
When should a headline be positive vs. negative?
Positive headlines work well for instructional content ("How to Double Your Traffic in 30 Days"). Negative headlines work well for problem-framing content that names a pain point ("5 Headline Mistakes Killing Your Click-Through Rate"). Both earn click-through because they trigger an emotional response; the choice depends on the content's angle. A neutral headline — flat, factual, no valence — is the weakest pattern for click-through, which is why Seeklyn's guidance for a neutral sentiment is: "A positive or negative angle tends to earn more clicks."
Frequently asked questions
Is a sentiment strength of 100 always good?
High strength means the headline has a strong emotional charge, which is generally good for click-through. However, very strong negative sentiment at scale can also read as alarmist. The sentiment sub-score rewards strength regardless of direction, so the signal is about emotional engagement, not about being positive.
Why does Seeklyn use AFINN instead of a machine-learning sentiment model?
AFINN is deterministic, interpretable, and fast — it produces the same score every time for the same input and tells you exactly which words drove the result. A neural sentiment model would be a black box. Seeklyn's design favours explainability so writers know what to change.
What if my headline has no AFINN words?
The net valence is 0, the polarity is neutral, and the strength is 0. The sentiment sub-score is 0, which reduces the CTR component. Adding 1 word with a strong AFINN score (e.g. "powerful" at +3) is often the lowest-effort way to lift the sentiment sub-score.