Plugin/new transliteration plugin (#215)

* add: Add TransliterationPlugin for changing english to hindi date-time mappings for deepgram ASR

* update: update packages for vuln fix

dialogy/plugins/registry.py

@@ -17,6 +17,7 @@
 from dialogy.plugins.text.oos_filter import OOSFilterPlugin
 from dialogy.plugins.text.intent_entity_mutator import IntentEntityMutatorPlugin
 from dialogy.plugins.text.qc_plugin import QCPlugin
+from dialogy.plugins.text.transliteration_plugin import TransliterationPlugin
 
 
 plugin_cls_lib = {
@@ -37,5 +38,6 @@
     "ErrorRecoveryPlugin": ErrorRecoveryPlugin,
     "OOSFilterPlugin": OOSFilterPlugin,
     "IntentEntityMutatorPlugin": IntentEntityMutatorPlugin,
-    "QCPlugin": QCPlugin
-}
+    "QCPlugin": QCPlugin,
+    "TransliterationPlugin": TransliterationPlugin
+}

dialogy/plugins/text/list_entity_plugin/__init__.py

@@ -173,14 +173,14 @@ def _parse(self, candidates: Optional[Dict[str, Dict[str, List[Any]]]]) -> None:
         :return: None
         :rtype: NoneType
         """
-        logger.debug(
-            pformat(
-                {
-                    "style": self.style,
-                    "candidates": candidates,
-                }
-            )
-        )
+        # logger.debug(
+        #     pformat(
+        #         {
+        #             "style": self.style,
+        #             "candidates": candidates,
+        #         }
+        #     )
+        # )
         if not isinstance(candidates, dict):
             raise TypeError(
                 'Expected "candidates" to be a Dict[str, List[str]]'
@@ -214,8 +214,8 @@ def _parse(self, candidates: Optional[Dict[str, Dict[str, List[Any]]]]) -> None:
                     else:
                         self.compiled_patterns[entity_type] = {entity_value: patterns}
 
-        logger.debug("compiled patterns")
-        logger.debug(self.compiled_patterns)
+        # logger.debug("compiled patterns")
+        # logger.debug(self.compiled_patterns)
 
     def _search(self, transcripts: List[str]) -> List[MatchType]:
         """

dialogy/plugins/text/transliteration_plugin/__init__.py

@@ -0,0 +1,153 @@
+"""
+This plugin transliterates text between languages, specifically handling date/time words from English to Hindi.
+It should be placed after the MergeASROutputPlugin in the pipeline.
+"""
+import string
+from typing import Any, Dict, List, Optional
+from loguru import logger
+
+import dialogy.constants as const
+from dialogy.base import Guard, Input, Output, Plugin
+
+
+class TransliterationPlugin(Plugin):
+    """
+    Transliterates English date/time words to Hindi to improve downstream processing.
+
+    The ASR transcripts may contain English words mixed with Hindi (code-mixing),
+    particularly for date and time references. This can cause issues with plugins
+    like Duckling that expect consistent language usage.
+
+    Example:
+        Input: "मैंने last week Tuesday को payment किया"
+        Output: "मैंने पिछला सप्ताह मंगलवार को payment किया"
+
+    This plugin identifies English date/time words and replaces them with Hindi translations
+    while preserving other words.
+    """
+    # Class-level constant mapping for better performance
+    DATETIME_MAPPING = {
+        # Days
+        "monday": "सोमवार",
+        "tuesday": "मंगलवार", 
+        "wednesday": "बुधवार",
+        "thursday": "गुरुवार",
+        "friday": "शुक्रवार",
+        "saturday": "शनिवार",
+        "sunday": "रविवार",
+
+        # Months  
+        "january": "जनवरी",
+        "february": "फरवरी", 
+        "march": "मार्च",
+        "april": "अप्रैल",
+        "may": "मई",
+        "june": "जून",
+        "july": "जुलाई", 
+        "august": "अगस्त",
+        "september": "सितंबर",
+        "october": "अक्तूबर",
+        "november": "नवंबर",
+        "december": "दिसंबर",
+
+        # Times of day
+        "morning": "सुबह",
+        "afternoon": "दोपहर", 
+        "evening": "शाम",
+        "night": "रात",
+        "am": "पूर्वाह्न",
+        "pm": "अपराह्न",
+
+        # Relative days
+        "today": "आज",
+        "tomorrow": "कल",
+        "yesterday": "कल",
+
+        # Time units
+        "hour": "घंटा",
+        "minute": "मिनट",
+        "second": "सेकंड",
+        "week": "सप्ताह",
+        "month": "महीना", 
+        "year": "वर्ष",
+
+        # Modifiers
+        "next": "अगला",
+        "last": "पिछला",
+        "first": "पहला"
+    }
+
+    def __init__(
+        self,
+        input_column: str = const.ALTERNATIVES,
+        output_column: Optional[str] = None,
+        use_transform: bool = False,
+        dest: Optional[str] = None,
+        guards: Optional[List[Guard]] = None,
+        debug: bool = False,
+        **kwargs: Any
+    ) -> None:
+        super().__init__(
+            dest=dest,
+            guards=guards,
+            input_column=input_column,
+            output_column=output_column,
+            use_transform=use_transform,
+            debug=debug,
+            **kwargs
+        )
+
+    def map_to_hindi(self, transcript: str) -> str:
+        """
+        Maps English date/time words to their Hindi equivalents.
+        
+        Args:
+            transcript: Input text containing mixed English-Hindi words
+            
+        Returns:
+            Text with English date/time words replaced by Hindi equivalents
+        """
+        # Remove punctuation and convert to lowercase for consistent matching
+        words = transcript.lower().translate(str.maketrans('', '', string.punctuation)).split()
+
+        # Replace words that have Hindi equivalents, keep others unchanged
+        translated_words = [
+            self.DATETIME_MAPPING.get(word, word)
+            for word in words
+        ]
+        return ' '.join(translated_words)
+
+    def modify_utterances(self, utterances: List[List[Dict[str, str]]]) -> List[List[Dict[str, str]]]:
+        """
+        Modifies utterances by transliterating date/time words.
+        
+        Args:
+            utterances: List of ASR transcripts with confidence scores
+            
+        Returns:
+            Modified utterances with transliterated text
+        """
+        if not utterances or not utterances[0]:
+            return utterances
+
+        for transcript in utterances[0]:
+            if 'transcript' in transcript:
+                transcript['transcript'] = self.map_to_hindi(transcript['transcript'])
+
+        return utterances
+
+    async def utility(self, input_: Input, output: Output) -> Any:
+        """
+        Plugin utility method that processes the input utterances.
+        
+        Args:
+            input_: Input object containing utterances
+            output: Output object for results
+            
+        Returns:
+            Modified utterances with transliterated text
+        """
+        logger.debug(f"Input utterances for transliteration:\n{input_.utterances}")
+        modified_utterances = self.modify_utterances(input_.utterances)
+        logger.debug(f"Transliterated utterances:\n{modified_utterances}")
+        return modified_utterances