Update app.py

app.py

@@ -1,3 +1,4 @@
+# Environment settings
 import os
 os.environ["HF_HOME"] = "/tmp"
 os.environ["TRANSFORMERS_CACHE"] = "/tmp"
@@ -35,6 +36,52 @@ number_words = {
     100: "boqol", 1000: "kun"
 }
 
+shortcut_map = {
+    "asc": "asalaamu caleykum",
+    "wcs": "wacaleykum salaam",
+    "fcn": "fiican",
+    "xld": "xaaladda ka waran",
+    "kwrn": "kawaran",
+    "scw": "salalaahu caleyhi wa salam",
+    "alx": "alxamdu lilaahi",
+    "m.a": "maasha allah",
+    "sthy": "side tahey",
+    "sxp": "saaxiib"
+}
+
+country_map = {
+    "somalia": "Soomaaliya",
+    "ethiopia": "Itoobiya",
+    "kenya": "Kenya",
+    "djibouti": "Jabuuti",
+    "sudan": "Suudaan",
+    "Yeman": "yemaan",
+    "uganda": "Ugaandha",
+    "tanzania": "Tansaaniya",
+    "egypt": "Masar",
+    "libya": "Liibiya",
+    "algeria": "Aljeeriya",
+    "morocco": "Morooko",
+    "tunisia": "Tuniisiya",
+    "eritrea": "Eriteriya",
+    "malawi": "Malaawi",
+    "English": "ingiriis",
+    "Spain": "isbeen",
+    "Brazil": "baraasiil",
+    "niger": "Niyjer",
+    "Italy": "itaaliya",
+    "united states": "Maraykanka",
+    "china": "Shiinaha",
+    "india": "Hindiya",
+    "russia": "Ruushka",
+    "Saudi Arabia": "Sucuudi Carabiya",
+    "germany": "Jarmalka",
+    "france": "Faransiiska",
+    "japan": "Jabaan",
+    "canada": "Kanada",
+    "australia": "Australia"
+}
+
 def number_to_words(number):
     number = int(number)
     if number < 20:
@@ -50,21 +97,13 @@ def number_to_words(number):
         return part
     elif number < 1000000:
         thousands, remainder = divmod(number, 1000)
-        words = []
-        if thousands == 1:
-            words.append("kun")
-        else:
-            words.append(number_to_words(thousands) + " kun")
+        words = [number_to_words(thousands) + " kun" if thousands > 1 else "kun"]
         if remainder:
             words.append("iyo " + number_to_words(remainder))
         return " ".join(words)
     elif number < 1000000000:
         millions, remainder = divmod(number, 1000000)
-        words = []
-        if millions == 1:
-            words.append("milyan")
-        else:
-            words.append(number_to_words(millions) + " milyan")
+        words = [number_to_words(millions) + " milyan" if millions > 1 else "milyan"]
         if remainder:
             words.append(number_to_words(remainder))
         return " ".join(words)
@@ -72,11 +111,27 @@ def number_to_words(number):
         return str(number)
 
 def normalize_text(text):
+    text = re.sub(r'(?i)(?<!\w)zamzam(?!\w)', 'samsam', text)
+
+    def replace_shortcuts(match):
+        word = match.group(0).lower()
+        return shortcut_map.get(word, word)
+    pattern = re.compile(r'\b(' + '|'.join(re.escape(k) for k in shortcut_map.keys()) + r')\b', re.IGNORECASE)
+    text = pattern.sub(replace_shortcuts, text)
+
+    def replace_countries(match):
+        word = match.group(0).lower()
+        return country_map.get(word, word)
+    country_pattern = re.compile(r'\b(' + '|'.join(re.escape(k) for k in country_map.keys()) + r')\b', re.IGNORECASE)
+    text = country_pattern.sub(replace_countries, text)
+
     text = re.sub(r'(\d{1,3})(,\d{3})+', lambda m: m.group(0).replace(",", ""), text)
     text = re.sub(r'\.\d+', '', text)
+
     def replace_num(match):
         return number_to_words(match.group())
     text = re.sub(r'\d+', replace_num, text)
+
     symbol_map = {
         '$': 'doolar',
         '=': 'egwal',
@@ -85,9 +140,13 @@ def normalize_text(text):
     }
     for sym, word in symbol_map.items():
         text = text.replace(sym, ' ' + word + ' ')
+
     text = text.replace("KH", "qa").replace("Z", "S")
     text = text.replace("SH", "SHa'a").replace("DH", "Dha'a")
-    text = text.replace("ZamZam", "SamSam")
+
+    if re.search(r'(?i)(zamzam|samsam)[\s\.,!?]*$', text.strip()):
+        text += " m"
+
     return text
 
 def waveform_to_wav_bytes(waveform: torch.Tensor, sample_rate: int = 22050) -> bytes: