Fix bibliography titlecase (in clean_bibliography.sh) and a few things in the bibliography itself

src/thesis/clean_bibliography.sh

@@ -19,6 +19,8 @@ with open(path) as f:
 doi_re = re.compile(r"doi\s*=\s*\{10\.48550/arXiv\.([^}]+)\}")
 type_re = re.compile(r"^@([A-Za-z]+)\{", re.MULTILINE)
 howpublished_re = re.compile(r"^\s*howpublished\s*=\s*\{", re.MULTILINE)
+title_field_re = re.compile(r"\b(title|booktitle)\s*=\s*\{", re.IGNORECASE)
+inner_brace_re = re.compile(r"\{([A-Za-z0-9]+)\}")
 
 # Split into entries by scanning for top-level "@type{...}" blocks. We walk
 # brace depth so that the closing "}" of the entry is matched correctly even
@@ -47,7 +49,7 @@ def split_entries(s):
         i = j
     return out
 
-def transform(entry):
+def normalize_arxiv(entry):
     doi_m = doi_re.search(entry)
     if not doi_m:
         return entry
@@ -64,6 +66,43 @@ def transform(entry):
         )
     return entry
 
+# Strip protective braces around words inside title/booktitle values.
+# BibTeX uses "{Word}" inside titles to preserve case against the bibliography
+# style's title-casing rules. We keep that protection only when every character
+# inside the braces is non-lowercase (e.g. acronyms like {NASA}); for ordinary
+# words like {Quantum} we drop the braces so the style's casing applies.
+def strip_title_braces(entry):
+    out, i, n = [], 0, len(entry)
+    while True:
+        m = title_field_re.search(entry, i)
+        if not m:
+            out.append(entry[i:])
+            break
+        out.append(entry[i:m.end()])
+        depth, j = 1, m.end()
+        while j < n and depth > 0:
+            c = entry[j]
+            if c == "{":
+                depth += 1
+            elif c == "}":
+                depth -= 1
+                if depth == 0:
+                    break
+            j += 1
+        value = entry[m.end():j]
+        cleaned = inner_brace_re.sub(
+            lambda mm: mm.group(1) if any(c.islower() for c in mm.group(1)) else mm.group(0),
+            value,
+        )
+        out.append(cleaned)
+        if j < n:
+            out.append(entry[j])
+        i = j + 1
+    return "".join(out)
+
+def transform(entry):
+    return strip_title_braces(normalize_arxiv(entry))
+
 parts = split_entries(text)
 new_text = "".join(transform(p) if kind == "entry" else p for kind, p in parts)