Upload README.md with huggingface_hub

README.md

@@ -12,16 +12,13 @@ tags:
 - korean
 - english
 - code-switching
-- forced-alignment
 ---
 
 # wav2vec2-xls-r-300m-korean-english
 
-A **jamo + Latin CTC acoustic model** fine-tuned from `facebook/wav2vec2-xls-r-300m`
+A **jamo + Latin CTC** acoustic model fine-tuned from `facebook/wav2vec2-xls-r-300m`
 for **Korean / English code-switched** speech. It emits per-frame character
-probabilities, so it works both as a character recognizer and (its validated use)
-as a **streaming CTC forced aligner** — pair the emissions with known text + a
-Viterbi pass to get per-character timestamps (~20 ms granularity).
+probabilities over a compact jamo + Latin vocabulary.
 
 ## Training
 
@@ -34,7 +31,7 @@ Viterbi pass to get per-character timestamps (~20 ms granularity).
 ## Important: jamo preprocessing
 
 The vocab is **jamo (decomposed Hangul) + Latin**, not raw syllables. You MUST
-decompose reference/target text the same way, and recompose model output:
+decompose target text the same way, and recompose model output:
 
 ```python
 import unicodedata
@@ -44,10 +41,10 @@ def from_jamo(t): # model output -> readable text
     return unicodedata.normalize("NFC", t)
 ```
 
-## Usage (recognition)
+## Usage
 
 ```python
-import torch, soundfile as sf, unicodedata
+import soundfile as sf, unicodedata
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 
 proc = Wav2Vec2Processor.from_pretrained("raniczan/wav2vec2-xls-r-300m-korean-english")
@@ -58,6 +55,3 @@ x = proc(wav, sampling_rate=16000, return_tensors="pt").input_values
 ids = model(x).logits.argmax(-1)[0]
 print(unicodedata.normalize("NFC", proc.decode(ids)))
 ```
-
-For **forced alignment**, run the same logits through `torchaudio.functional.forced_align`
-(or a custom streaming Viterbi) against your jamo-decomposed known text.