Bayesian filter
ما هو Bayesian filter?
هذا المصطلح متوفر حاليًا باللغة الإنجليزية فقط.
A Bayesian filter is a spam filter that classifies a message by applying Bayes' theorem to the words and phrases it contains. The filter is trained on examples of known spam and known legitimate mail, so each token carries a spam probability — and the message's overall spam score comes from combining the probabilities of every token in it. Because the model learns from real examples, it adapts to new spam wording instead of depending on a hardcoded keyword list that senders can simply spell around.
More detail
Training splits the corpus in two: tokens that mostly appear in spam (prize claims, urgency phrasing, an actual sender's recurring scam copy) pull a message toward the spam folder, while tokens that mostly appear in the recipient's legitimate mail pull it toward the inbox. A well-trained Bayesian filter therefore reflects the specific mail that individual or organization actually receives, rather than one global definition of spam.
Bayesian classification is probabilistic, so its output is a score, not a verdict — receiving systems combine it with authentication results (SPF, DKIM, DMARC), sender reputation, and other heuristics before deciding placement. Spammers attempt to defuse the token math by seeding messages with long runs of innocent words, a technique known as Bayesian poisoning, which is why modern spam filters treat the Bayesian score as one signal among several.
الأسئلة الشائعة
- How does a Bayesian filter decide a message is spam?
- Each word or phrase in a message carries a spam probability learned from training examples; the filter combines those probabilities with Bayes' theorem into an overall spam score, and messages whose score passes the threshold are treated as spam.
- Why is Bayesian filtering better than a keyword blocklist?
- Because it learns from real examples instead of matching a fixed keyword list — a spammer who avoids the exact blocklisted words still gets caught when the overall token statistics read like spam, and the filter keeps improving as it sees the recipient's actual mix of spam and legitimate mail.
- What do senders need to know about Bayesian filters?
- Copy phrasing is one input into the spam score, so recycled scam-style phrasing hurts while the recurring language of genuinely solicited mail gets learned as legitimate. The heavier deliverability levers stay list hygiene, complaint rate, and authentication — a Bayesian score is only one signal a receiving system folds into the placement decision.
See also
Build it on Orbit
Voice, messaging, email, video, and AI agents on one platform and one pay-as-you-go bill. Start free — no credit card required.