From 99447e00a1419c6e9ef5bb3ffcbc6e9cde7f5eb3 Mon Sep 17 00:00:00 2001
From: dualfroz <me@dualfroz.com>
Date: Fri, 2 Oct 2026 16:06:32 +0200
Subject: [PATCH] Don't read a titlecase gerund after a colon as a proper noun

---
 .../preTagger/compute/tagger/2nd-pass/01-case.js   |  8 +++++++-
 src/2-two/preTagger/compute/tagger/index.js        |  9 ++++++---
 tests/two/tagger/ambiguity/gerund.test.js          | 14 ++++++++++++++
 3 files changed, 27 insertions(+), 4 deletions(-)

diff --git a/src/2-two/preTagger/compute/tagger/2nd-pass/01-case.js b/src/2-two/preTagger/compute/tagger/2nd-pass/01-case.js
index 3cf5379..c43adcf 100644
--- a/src/2-two/preTagger/compute/tagger/2nd-pass/01-case.js
+++ b/src/2-two/preTagger/compute/tagger/2nd-pass/01-case.js
@@ -20,7 +20,7 @@ const nope = {
 }
 
 // if it's a unknown titlecase word, it's a propernoun
-const checkCase = function (terms, i, model) {
+const checkCase = function (terms, i, model, afterColon) {
   const term = terms[i]
   // assume terms are already indexed
   term.index ||= [0, 0]
@@ -40,6 +40,12 @@ const checkCase = function (terms, i, model) {
     if (/["']$/.test(term.pre)) {
       return null
     }
+    // gerund after a colon - 'Tutorial: Creating a cake'
+    const before = terms[i - 1]
+    const hasColon = before ? /:/.test(before.post) : afterColon
+    if (hasColon && term.tags.has('Gerund')) {
+      return null
+    }
     if (term.normal === 'the') {
       return null
     }
diff --git a/src/2-two/preTagger/compute/tagger/index.js b/src/2-two/preTagger/compute/tagger/index.js
index 52dcc05..8a39e33 100644
--- a/src/2-two/preTagger/compute/tagger/index.js
+++ b/src/2-two/preTagger/compute/tagger/index.js
@@ -37,7 +37,7 @@ const firstPass = function (docs, model, world) {
 }
 
 // these methods don't care about word-neighbours
-const secondPass = function (terms, model, world, isYelling) {
+const secondPass = function (terms, model, world, isYelling, afterColon) {
   for (let i = 0; i < terms.length; i += 1) {
     // skip frozen terms, for now
     if (terms[i].frozen === true) {
@@ -47,7 +47,7 @@ const secondPass = function (terms, model, world, isYelling) {
     tagSwitch(terms, i, model)
     //  is it titlecased?
     if (isYelling === false) {
-      checkCase(terms, i, model)
+      checkCase(terms, i, model, afterColon)
     }
     // look at word ending
     checkSuffix(terms, i, model)
@@ -104,8 +104,11 @@ const preTagger = function (view) {
     const terms = document[n]
     // is it all upper-case?
     const isYelling = ignoreCase(terms)
+    // does it follow a colon?
+    const before = document[n - 1]
+    const afterColon = before !== undefined && /:/.test(before[before.length - 1].post)
     // guess by the letters
-    secondPass(terms, model, world, isYelling)
+    secondPass(terms, model, world, isYelling, afterColon)
     // guess by the neighbours
     thirdPass(terms, model, world, isYelling)
   }
diff --git a/tests/two/tagger/ambiguity/gerund.test.js b/tests/two/tagger/ambiguity/gerund.test.js
index 09c50ce..83da321 100644
--- a/tests/two/tagger/ambiguity/gerund.test.js
+++ b/tests/two/tagger/ambiguity/gerund.test.js
@@ -119,3 +119,17 @@ test('match:', function (t) {
   })
   t.end()
 })
+
+test('titlecase after a colon', function (t) {
+  const arr = [
+    ['Tutorial: Creating a cake', '. #Gerund #Determiner #Noun'],
+    ['Step two: Mixing the flour', '. #Value #Gerund the #Noun'],
+    ['On Tuesday: Baking a cake', 'on #Date #Singular a #Noun'],
+  ]
+  arr.forEach(function (a) {
+    const [str, match] = a
+    const doc = nlp(str)
+    t.equal(doc.match(match).text(), doc.text(), here + str)
+  })
+  t.end()
+})