nkululeko 0.95.8__tar.gz → 0.95.9__tar.gz

{nkululeko-0.95.8 → nkululeko-0.95.9}/CHANGELOG.md

@@ -1,6 +1,10 @@
 Changelog
 =========
 
+Version 0.95.9 (25-07-14)
+--------------------------
+* added google translation
+
 Version 0.95.8 (25-07-14)
 --------------------------
 * fix bug that bool was not detected as categorical

{nkululeko-0.95.8/nkululeko.egg-info → nkululeko-0.95.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nkululeko
-Version: 0.95.8
+Version: 0.95.9
 Summary: Machine learning audio prediction experiments based on templates
 Home-page: https://github.com/felixbur/nkululeko
 Author: Felix Burkhardt

nkululeko-0.95.9/nkululeko/autopredict/ap_translate.py

@@ -0,0 +1,39 @@
+"""A translator for text.
+
+Currently based on google translate.
+"""
+
+from nkululeko.utils.util import Util
+
+
+class TextTranslator:
+    """Translator.
+
+    translate text with the google translate model
+    """
+
+    def __init__(self, df, util=None):
+        self.df = df
+        if util is not None:
+            self.util = util
+        else:
+            # create a new util instance
+            # this is needed to access the config and other utilities
+            # in the autopredict module
+            self.util = Util("translator")
+            
+        self.language = self.util.config_val("PREDICT", "target_language", "en")
+        from nkululeko.autopredict.google_translator import GoogleTranslator
+        self.translator = GoogleTranslator(
+            language=self.language,
+            util=self.util,
+        )
+
+    def predict(self, split_selection):
+        self.util.debug(f"translating text for {split_selection} samples")
+        df = self.translator.translate_index(
+            self.df
+        )
+        return_df = self.df.copy()
+        return_df[self.language] = df[self.language].values
+        return return_df

nkululeko-0.95.9/nkululeko/autopredict/google_translator.py

@@ -0,0 +1,63 @@
+import os
+
+import pandas as pd
+import torch
+from tqdm import tqdm
+
+import asyncio
+from googletrans import Translator
+
+import audeer
+import audiofile
+
+from nkululeko.utils.util import Util
+
+import httpx
+
+class GoogleTranslator:
+    def __init__(self, language="en", util=None):
+        self.language = language
+        self.util = util
+
+    async def translate_text(self, text):
+        async with Translator() as translator:
+            result = translator.translate(text, dest="en")
+            return (await result).text
+
+    def translate_index(self, df:pd.DataFrame) ->  pd.DataFrame:
+        """Transcribe the audio files in the given index.
+
+        :param index: Index containing tuples of (file, start, end).
+        :return: DataFrame with transcriptions indexed by the original index.
+        :rtype: pd.DataFrame
+        """
+        file_name = ""
+        seg_index = 0
+        translations = []
+        translator_cache = audeer.mkdir(
+            audeer.path(self.util.get_path("cache"), "translations"))
+        file_name = ""
+        for idx, row in tqdm(df.iterrows(), total=len(df)):
+            file = idx[0]
+            start = idx[1]
+            end = idx[2]
+            if file != file_name:
+                file_name = file
+                seg_index = 0
+            cache_name = audeer.basename_wo_ext(file)+str(seg_index)
+            cache_path = audeer.path(translator_cache, cache_name + ".json")
+            if os.path.isfile(cache_path):
+                translation = self.util.read_json(cache_path)["translation"]
+            else:
+                text = row['text']
+                translation = asyncio.run(self.translate_text(text))
+                self.util.save_json(cache_path, 
+                                {"translation": translation, 
+                                 "file": file, 
+                                 "start": start.total_seconds(), 
+                                 "end": end.total_seconds()})
+            translations.append(translation)
+            seg_index += 1
+
+        df = pd.DataFrame({self.language:translations}, index=df.index)
+        return df

{nkululeko-0.95.8 → nkululeko-0.95.9}/nkululeko/constants.py

@@ -1,2 +1,2 @@
-VERSION="0.95.8"
+VERSION="0.95.9"
 SAMPLING_RATE = 16000

{nkululeko-0.95.8 → nkululeko-0.95.9}/nkululeko/experiment.py

@@ -574,6 +574,11 @@ class Experiment:
 
                 predictor = TextPredictor(df, self.util)
                 df = predictor.predict(sample_selection)
+            elif target == "translation":
+                from nkululeko.autopredict.ap_translate import TextTranslator
+
+                predictor = TextTranslator(df, self.util)
+                df = predictor.predict(sample_selection)
             elif target == "arousal":
                 from nkululeko.autopredict.ap_arousal import ArousalPredictor
 

{nkululeko-0.95.8 → nkululeko-0.95.9/nkululeko.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nkululeko
-Version: 0.95.8
+Version: 0.95.9
 Summary: Machine learning audio prediction experiments based on templates
 Home-page: https://github.com/felixbur/nkululeko
 Author: Felix Burkhardt

{nkululeko-0.95.8 → nkululeko-0.95.9}/nkululeko.egg-info/SOURCES.txt

@@ -108,8 +108,10 @@ nkululeko/autopredict/ap_sid.py
 nkululeko/autopredict/ap_snr.py
 nkululeko/autopredict/ap_stoi.py
 nkululeko/autopredict/ap_text.py
+nkululeko/autopredict/ap_translate.py
 nkululeko/autopredict/ap_valence.py
 nkululeko/autopredict/estimate_snr.py
+nkululeko/autopredict/google_translator.py
 nkululeko/autopredict/whisper_transcriber.py
 nkululeko/autopredict/tests/__init__.py
 nkululeko/autopredict/tests/test_whisper_transcriber.py

{nkululeko-0.95.8 → nkululeko-0.95.9}/pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "nkululeko"
-version = "0.95.8"
+version = "0.95.9"
 description = "Machine learning audio prediction experiments based on templates"
 authors = [
     {name = "Felix Burkhardt", email = "fxburk@gmail.com"},