added character substitution

README.md

@@ -22,7 +22,7 @@ model-index:
     metrics:
        - name: Test WER
          type: wer
-         value: 29.48 
+         value: 29.35 
 ---
 
 # Wav2Vec2-Large-XLSR-53-German
@@ -88,14 +88,37 @@ model = Wav2Vec2ForCTC.from_pretrained("de")
 `elgeish/wav2vec2-large-xlsr-53-arabic`
 model.to("cuda")
 
-chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\”\�\カ\æ\無\ན\カ\臣\ѹ\…\«\»\ð\ı\„\幺\א\ב\比\ш\ע\)\ứ\в\œ\ч\+\—\ш\‚\נ\м\ń\乡\$\=\ש\ф\支\(\°\и\к\̇]'  
+chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\”\�\カ\æ\無\ན\カ\臣\ѹ\…\«\»\ð\ı\„\幺\א\ב\比\ш\ע\)\ứ\в\œ\ч\+\—\ш\‚\נ\м\ń\乡\$\=\ש\ф\支\(\°\и\к\̇]'
+substitutions = {
+\t'e' : '[\ə\é\ě\ę\ê\ế\ế\ë\ė\е]',
+\t'o' : '[\ō\ô\ô\ó\ò\ø\ọ\ŏ\õ\ő\о]',
+\t'a' : '[\á\ā\ā\ă\ã\å\â\à\ą\а]',
+\t'c' : '[\č\ć\ç\с]',
+\t'l' : '[\ł]',
+\t'u' : '[\ú\ū\ứ\ů]',
+\t'und' : '[\&]',
+\t'r' : '[\ř]',
+\t'y' : '[\ý]',
+\t's' : '[\ś\š\ș\ş]',
+\t'i' : '[\ī\ǐ\í\ï\î\ï]',
+\t'z' : '[\ź\ž\ź\ż]',
+\t'n' : '[\ñ\ń\ņ]',
+\t'g' : '[\ğ]',
+\t'ss' : '[\ß]',
+\t't' : '[\ț\ť]',
+\t'd' : '[\ď\đ]',
+\t"'": '[\ʿ\་\’\`\´\ʻ\`\‘]',
+\t'p': '\р'
+}
 resampler = torchaudio.transforms.Resample(48_000, 16_000)
 
 # Preprocessing the datasets.
 # We need to read the aduio files as arrays
 def speech_file_to_array_fn(batch):
 \tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
-\tbatch["sentence"] = re.sub('\ß', 'ss', batch["sentence"])
+\tfor x in substitutions:
+\t\tbatch["sentence"] = re.sub(substitutions[x], x, batch["sentence"])
+\t\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
 \tspeech_array, sampling_rate = torchaudio.load(batch["path"])
 \tbatch["speech"] = resampler(speech_array).squeeze().numpy()
 \treturn batch
@@ -119,7 +142,7 @@ result = test_dataset.map(evaluate, batched=True, batch_size=8)
 print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
 ```
 
-**Test Result**: 29.48 % 
+**Test Result**: 29.35 % 
 
 
 ## Training