From 68919ed87aa6a097a4d21414ec7c438883f5fe22 Mon Sep 17 00:00:00 2001
From: NetterSilver <133535268+Silvertree2010@users.noreply.github.com>
Date: Mon, 2 Nov 2026 06:38:18 +0100
Subject: [PATCH] Read decades like 1990s as plural years in TTS normalizer
 (#3930)

* Read decades like 1990s as plural years in TTS normalizer

Found while testing crates/tts-text/src/normalize.rs:~654 (cardinal/year branch) and parse_continuation ~line 680

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* tts-text: decade tests in the it binary, with boundary cases

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Haiku 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
---
 crates/tts-text/src/normalize.rs    | 22 ++++++++++++++++++++++
 crates/tts-text/tests/it/decades.rs | 27 +++++++++++++++++++++++++++
 crates/tts-text/tests/it/main.rs    |  1 +
 3 files changed, 50 insertions(+)
 create mode 100644 crates/tts-text/tests/it/decades.rs
 create mode 100644 crates/tts-text/tests/it/main.rs

diff --git a/crates/tts-text/src/normalize.rs b/crates/tts-text/src/normalize.rs
index d925b086b..834a9c01a 100644
--- a/crates/tts-text/src/normalize.rs
+++ b/crates/tts-text/src/normalize.rs
@@ -654,9 +654,31 @@ fn parse_number_core(rest: &str, lang: Lang) -> Option<(Vec<Token>, usize)> {
     if !grouped && digits.len() == 4 && (1100..=2099).contains(&value) && year_is_standalone(rest, used) {
         return Some((year_words(value, lang), used));
     }
+    // A decade such as "1990s" reads as a plural year: "nineteen nineties".
+    if !grouped
+        && digits.len() == 4
+        && (1100..=2099).contains(&value)
+        && value.is_multiple_of(10)
+        && matches!(peek(rest, used), Some('s'))
+        && !peek(rest, used + 1).is_some_and(char::is_alphanumeric)
+    {
+        let mut words = year_words(value, lang);
+        if let Some(Token::Word(last)) = words.last_mut() {
+            *last = pluralize_decade(last);
+        }
+        return Some((words, used + 1));
+    }
     Some((english_cardinal(value).into_iter().map(Token::Word).collect(), used))
 }
 
+/// Plural of the last word of a round year: ninety → nineties, hundred → hundreds.
+fn pluralize_decade(word: &str) -> String {
+    match word.strip_suffix('y') {
+        Some(stem) => format!("{stem}ies"),
+        None => format!("{word}s"),
+    }
+}
+
 fn has_leading_zero_str(digits: &str) -> bool {
     digits.len() > 1 && digits.starts_with('0')
 }
diff --git a/crates/tts-text/tests/it/decades.rs b/crates/tts-text/tests/it/decades.rs
new file mode 100644
index 000000000..1316bcf4e
--- /dev/null
+++ b/crates/tts-text/tests/it/decades.rs
@@ -0,0 +1,27 @@
+use filmcraft_tts_text::{Lang, Token, normalize};
+
+fn spoken(text: &str) -> String {
+    let words: Vec<String> = normalize(text, Lang::EnUs).unwrap().into_iter().filter_map(|t| if let Token::Word(w) = t { Some(w) } else { None }).collect();
+    words.join(" ")
+}
+
+#[test]
+fn bare_year_is_unchanged() {
+    assert_eq!(spoken("in 1990"), "in nineteen ninety");
+}
+
+#[test]
+fn decade_reads_as_plural_year() {
+    assert_eq!(spoken("the 1990s"), "the nineteen nineties");
+    assert_eq!(spoken("the 2010s."), "the twenty tens");
+    assert_eq!(spoken("the 1900s"), "the nineteen hundreds");
+    assert_eq!(spoken("the 1100s"), "the eleven hundreds");
+    assert_eq!(spoken("the 2090s"), "the twenty nineties");
+}
+
+#[test]
+fn only_round_years_followed_by_a_lone_s_are_decades() {
+    assert_ne!(spoken("the 1995s"), "the nineteen ninety fives");
+    assert!(!spoken("the 1990sx").contains("nineties"));
+    assert!(!spoken("the 3000s").contains("thousands"));
+}
diff --git a/crates/tts-text/tests/it/main.rs b/crates/tts-text/tests/it/main.rs
new file mode 100644
index 000000000..f45cbb417
--- /dev/null
+++ b/crates/tts-text/tests/it/main.rs
@@ -0,0 +1 @@
+mod decades;