dev day #79

internal/services/spam.go

@@ -0,0 +1,55 @@
+package services
+
+import (
+	"regexp"
+	"strings"
+)
+
+// Simple heuristics to evaluate spammy text. Returns score 0..1 and triggered rules.
+func EvaluateSpamScore(s string) (float64, []string) {
+	var rules []string
+	content := strings.TrimSpace(s)
+	if content == "" {
+		return 1.0, []string{"empty"}
+	}
+	// Too short
+	if len([]rune(content)) < 6 {
+		rules = append(rules, "too_short")
+	}
+	// Excessive repeated characters like 'aaaaaa' or '!!!!'
+	repeatRe := regexp.MustCompile(`([a-zA-Z!?.])\1{4,}`)
+	if repeatRe.MatchString(content) {
+		rules = append(rules, "repeated_chars")
+	}
+	// Low vowel ratio suggests gibberish in Czech/English latin text
+	letters := regexp.MustCompile(`[A-Za-zÁáÉéĚěÍíÓóÚúŮůÝýŽžŠšČčŘřŤťŇňĎď]`).FindAllString(content, -1)
+	if len(letters) >= 8 {
+		vowels := regexp.MustCompile(`[AaEeIiOoUuYyÁáÉéĚěÍíÓóÚúŮůÝý]`).FindAllString(content, -1)
+		ratio := float64(len(vowels)) / float64(len(letters))
+		if ratio < 0.18 { // very low vowel ratio
+			rules = append(rules, "low_vowel_ratio")
+		}
+	}
+	// Too many links
+	linkCount := len(regexp.MustCompile(`https?://`).FindAllStringIndex(content, -1))
+	if linkCount >= 3 {
+		rules = append(rules, "too_many_links")
+	}
+	// All-caps shouting
+	if content == strings.ToUpper(content) && len(content) >= 8 {
+		rules = append(rules, "all_caps")
+	}
+	// Compute score by rules weight
+	weights := map[string]float64{
+		"empty":           1.0,
+		"too_short":       0.4,
+		"repeated_chars":  0.3,
+		"low_vowel_ratio": 0.3,
+		"too_many_links":  0.5,
+		"all_caps":        0.2,
+	}
+	score := 0.0
+	for _, r := range rules { score += weights[r] }
+	if score > 1.0 { score = 1.0 }
+	return score, rules
+}