Vadym Myroshnyk commited on
Commit ·
edf0091
1
Parent(s): 2e3d053
save
Browse files- app.py +37 -23
- audio_files.py +2 -3
- check.py +7 -9
app.py
CHANGED
|
@@ -1,10 +1,11 @@
|
|
| 1 |
import os
|
| 2 |
import gradio as gr
|
|
|
|
| 3 |
|
| 4 |
from audio_files import get_audio_structure
|
| 5 |
-
from check import
|
| 6 |
|
| 7 |
-
# === Audio structure
|
| 8 |
audio_structure = get_audio_structure()
|
| 9 |
week_options = list(audio_structure.keys())
|
| 10 |
default_week = week_options[0]
|
|
@@ -13,55 +14,68 @@ default_file = audio_structure[default_week][0] if audio_structure[default_week]
|
|
| 13 |
with gr.Blocks(theme="soft") as demo:
|
| 14 |
gr.Markdown("## 🎧 Modals & Conditionals Bootcamp")
|
| 15 |
|
| 16 |
-
# States
|
| 17 |
current_index = gr.State(0)
|
| 18 |
dialogue_state = gr.State([])
|
| 19 |
|
| 20 |
-
# Week + audio
|
| 21 |
with gr.Row():
|
| 22 |
week_dropdown = gr.Dropdown(label="Select a week", choices=week_options, value=default_week, scale=1)
|
| 23 |
audio_player = gr.Audio(label="Audio preview", type="filepath", value=default_file, scale=3)
|
| 24 |
|
| 25 |
gr.Markdown("### ✍️ Your Transcription")
|
| 26 |
|
| 27 |
-
# === Dynamic dialogue render ===
|
| 28 |
@gr.render(inputs=[week_dropdown, current_index, dialogue_state])
|
| 29 |
def render_transcription_inputs(week, index, inputs):
|
| 30 |
audio_path = audio_structure[week][0]
|
| 31 |
transcript_path = audio_path.replace(".mp3", ".txt").replace("audio", "transcripts")
|
| 32 |
lines = load_dialogue_lines(transcript_path) if os.path.exists(transcript_path) else []
|
| 33 |
|
| 34 |
-
# Add next speaker
|
| 35 |
if index < len(lines):
|
| 36 |
next_speaker, _ = lines[index]
|
| 37 |
inputs = [(next_speaker, "")] + inputs
|
| 38 |
index += 1
|
| 39 |
|
| 40 |
-
# === UI Layout ===
|
| 41 |
button_label = "➕ Add next speaker" if index < len(lines) else "✅ Check"
|
| 42 |
action_btn = gr.Button(button_label)
|
| 43 |
-
textboxes = [gr.Textbox(label=speaker, value=value, lines=1, key=f"{week}-{i}") for i, (speaker, value) in enumerate(inputs)]
|
| 44 |
|
| 45 |
-
|
| 46 |
-
|
|
|
|
| 47 |
|
| 48 |
-
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
-
# === Event binding (depending on mode)
|
| 60 |
if index < len(lines):
|
| 61 |
-
action_btn.click(fn=
|
| 62 |
else:
|
| 63 |
-
action_btn.click(
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
demo.load(lambda: None)
|
| 67 |
|
|
|
|
| 1 |
import os
|
| 2 |
import gradio as gr
|
| 3 |
+
from rapidfuzz import fuzz
|
| 4 |
|
| 5 |
from audio_files import get_audio_structure
|
| 6 |
+
from check import load_dialogue_lines, highlight_fuzzy_diff # додано highlight_fuzzy_diff
|
| 7 |
|
| 8 |
+
# === Audio structure ===
|
| 9 |
audio_structure = get_audio_structure()
|
| 10 |
week_options = list(audio_structure.keys())
|
| 11 |
default_week = week_options[0]
|
|
|
|
| 14 |
with gr.Blocks(theme="soft") as demo:
|
| 15 |
gr.Markdown("## 🎧 Modals & Conditionals Bootcamp")
|
| 16 |
|
|
|
|
| 17 |
current_index = gr.State(0)
|
| 18 |
dialogue_state = gr.State([])
|
| 19 |
|
|
|
|
| 20 |
with gr.Row():
|
| 21 |
week_dropdown = gr.Dropdown(label="Select a week", choices=week_options, value=default_week, scale=1)
|
| 22 |
audio_player = gr.Audio(label="Audio preview", type="filepath", value=default_file, scale=3)
|
| 23 |
|
| 24 |
gr.Markdown("### ✍️ Your Transcription")
|
| 25 |
|
|
|
|
| 26 |
@gr.render(inputs=[week_dropdown, current_index, dialogue_state])
|
| 27 |
def render_transcription_inputs(week, index, inputs):
|
| 28 |
audio_path = audio_structure[week][0]
|
| 29 |
transcript_path = audio_path.replace(".mp3", ".txt").replace("audio", "transcripts")
|
| 30 |
lines = load_dialogue_lines(transcript_path) if os.path.exists(transcript_path) else []
|
| 31 |
|
|
|
|
| 32 |
if index < len(lines):
|
| 33 |
next_speaker, _ = lines[index]
|
| 34 |
inputs = [(next_speaker, "")] + inputs
|
| 35 |
index += 1
|
| 36 |
|
|
|
|
| 37 |
button_label = "➕ Add next speaker" if index < len(lines) else "✅ Check"
|
| 38 |
action_btn = gr.Button(button_label)
|
|
|
|
| 39 |
|
| 40 |
+
textboxes = []
|
| 41 |
+
feedback_outputs = []
|
| 42 |
+
diff_outputs = []
|
| 43 |
|
| 44 |
+
for i, (speaker, value) in enumerate(inputs):
|
| 45 |
+
with gr.Column():
|
| 46 |
+
tb = gr.Textbox(label=speaker, value=value, lines=1, key=f"{week}-{i}")
|
| 47 |
+
diff = gr.HTML()
|
| 48 |
+
fb = gr.HTML()
|
| 49 |
+
textboxes.append(tb)
|
| 50 |
+
diff_outputs.append(diff)
|
| 51 |
+
feedback_outputs.append(fb)
|
| 52 |
|
| 53 |
+
def on_add():
|
| 54 |
+
return index, inputs
|
| 55 |
+
|
| 56 |
+
def per_line_check(*user_inputs):
|
| 57 |
+
feedback = []
|
| 58 |
+
highlights = []
|
| 59 |
+
for user_input, (_, expected) in zip(user_inputs, reversed(lines[:len(user_inputs)])):
|
| 60 |
+
score = fuzz.ratio(user_input.strip(), expected.strip())
|
| 61 |
+
if score > 90:
|
| 62 |
+
emoji, msg = "✅", f"<b>{score}%</b> Excellent!"
|
| 63 |
+
elif score > 70:
|
| 64 |
+
emoji, msg = "⚠️", f"<b>{score}%</b> Some mistakes"
|
| 65 |
+
else:
|
| 66 |
+
emoji, msg = "❌", f"<b>{score}%</b> Try again"
|
| 67 |
+
feedback.append(f"<div style='margin-top: 4px;'>{emoji} {msg}</div>")
|
| 68 |
+
highlights.append(highlight_fuzzy_diff(user_input, expected))
|
| 69 |
+
return highlights + feedback # concatenate results
|
| 70 |
|
|
|
|
| 71 |
if index < len(lines):
|
| 72 |
+
action_btn.click(fn=on_add, outputs=[current_index, dialogue_state])
|
| 73 |
else:
|
| 74 |
+
action_btn.click(
|
| 75 |
+
fn=per_line_check,
|
| 76 |
+
inputs=textboxes,
|
| 77 |
+
outputs=diff_outputs + feedback_outputs
|
| 78 |
+
)
|
| 79 |
|
| 80 |
demo.load(lambda: None)
|
| 81 |
|
audio_files.py
CHANGED
|
@@ -4,8 +4,7 @@ import os
|
|
| 4 |
def get_audio_structure(audio_dir="audio"):
|
| 5 |
structure = {}
|
| 6 |
for file in sorted(os.listdir(audio_dir)):
|
| 7 |
-
if file.
|
| 8 |
-
label =
|
| 9 |
structure[label] = [os.path.join(audio_dir, file)]
|
| 10 |
return structure
|
| 11 |
-
|
|
|
|
| 4 |
def get_audio_structure(audio_dir="audio"):
|
| 5 |
structure = {}
|
| 6 |
for file in sorted(os.listdir(audio_dir)):
|
| 7 |
+
if file.endswith(".mp3"):
|
| 8 |
+
label = os.path.splitext(file)[0].replace("_", " ").title()
|
| 9 |
structure[label] = [os.path.join(audio_dir, file)]
|
| 10 |
return structure
|
|
|
check.py
CHANGED
|
@@ -60,12 +60,10 @@ def check_transcription(audio_path, user_text):
|
|
| 60 |
|
| 61 |
|
| 62 |
def load_dialogue_lines(path):
|
| 63 |
-
with open(path, "r") as f:
|
| 64 |
-
lines =
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
dialogue.append((speaker.strip(), text.strip()))
|
| 71 |
-
return dialogue
|
|
|
|
| 60 |
|
| 61 |
|
| 62 |
def load_dialogue_lines(path):
|
| 63 |
+
with open(path, "r", encoding="utf-8") as f:
|
| 64 |
+
lines = []
|
| 65 |
+
for line in f:
|
| 66 |
+
if ":" in line:
|
| 67 |
+
speaker, text = line.strip().split(":", 1)
|
| 68 |
+
lines.append((speaker.strip(), text.strip()))
|
| 69 |
+
return lines
|
|
|
|
|
|