Fix device type inference and add computer category

tools/device_mapper.py

@@ -177,11 +177,17 @@ WEAR_KEYWORDS = [
     "glasses",
     "眼镜",
 ]
-OTHER_KEYWORDS = [
+COMPUTER_KEYWORDS = [
     "matebook",
+    "macbook",
     "笔记本",
+    "电脑",
     "laptop",
     "notebook",
+    "desktop",
+    "workstation",
+]
+OTHER_KEYWORDS = [
     "vision",
     "vr",
     "ipod",
@@ -248,9 +254,12 @@ def brand_aliases(brand: str) -> List[str]:
 
 
 def has_keyword(text: str, keywords: Iterable[str]) -> bool:
-    norm_text = normalize_text(text)
+    norm_text = re.sub(r"[^0-9a-z\u4e00-\u9fff]+", " ", text.lower())
+    norm_text = " ".join(norm_text.split())
     for kw in keywords:
-        if normalize_text(kw) and normalize_text(kw) in norm_text:
+        kw_norm = re.sub(r"[^0-9a-z\u4e00-\u9fff]+", " ", kw.lower())
+        kw_norm = " ".join(kw_norm.split())
+        if kw_norm and kw_norm in norm_text:
             return True
     return False
 
@@ -300,6 +309,8 @@ def infer_device_type(
         return "tablet"
     if has_keyword(corpus, WEAR_KEYWORDS):
         return "wear"
+    if has_keyword(corpus, COMPUTER_KEYWORDS):
+        return "computer"
     if has_keyword(corpus, OTHER_KEYWORDS):
         return "other"
     if has_keyword(corpus, PHONE_KEYWORDS):
@@ -385,6 +396,7 @@ def parse_brand_file(path: Path) -> List[DeviceRecord]:
 
         section_match = SECTION_RE.match(line)
         if section_match:
+            flush_current()
             section = section_match.group(1).strip()
             continue