Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import speech_recognition as sr
|
3 |
from Levenshtein import ratio
|
@@ -9,24 +10,25 @@ import pandas as pd
|
|
9 |
# Sample dataframe with sentences
|
10 |
data = {
|
11 |
"Sentences": [
|
12 |
-
"
|
13 |
-
"An apple a day keeps the doctor away.",
|
14 |
"To be or not to be, that is the question.",
|
15 |
-
"
|
|
|
|
|
16 |
"She sells sea shells by the sea shore.",
|
17 |
-
"
|
18 |
-
"A stitch in time saves nine.",
|
19 |
-
"Good things come to those who wait.",
|
20 |
"Time flies like an arrow; fruit flies like a banana.",
|
21 |
-
"
|
|
|
22 |
]
|
23 |
}
|
24 |
df = pd.DataFrame(data)
|
|
|
25 |
|
26 |
def transcribe_audio(file_info):
|
27 |
r = sr.Recognizer()
|
28 |
with tempfile.NamedTemporaryFile(delete=True, suffix=".wav") as tmpfile:
|
29 |
-
sf.write(
|
30 |
tmpfile.seek(0)
|
31 |
with sr.AudioFile(tmpfile.name) as source:
|
32 |
audio_data = r.record(source)
|
@@ -38,36 +40,50 @@ def transcribe_audio(file_info):
|
|
38 |
except sr.RequestError as e:
|
39 |
return f"Could not request results; {e}"
|
40 |
|
41 |
-
def pronunciation_correction(
|
42 |
-
expected_text = selected_sentence
|
43 |
user_spoken_text = transcribe_audio(file_info)
|
44 |
similarity = ratio(expected_text.lower(), user_spoken_text.lower())
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
feedback = "Excellent pronunciation!"
|
49 |
-
elif similarity >= 0.7:
|
50 |
-
feedback = "Good pronunciation!"
|
51 |
-
elif similarity >= 0.5:
|
52 |
-
feedback = "Needs improvement."
|
53 |
else:
|
54 |
-
|
55 |
-
|
56 |
-
|
|
|
|
|
|
|
57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
gr.Dropdown(choices=df['Sentences'].tolist(), label="Select a Sentence")
|
63 |
-
gr.
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
)
|
71 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
|
73 |
-
|
|
|
1 |
+
#@markdown Accuracy Score, Average, user name
|
2 |
import gradio as gr
|
3 |
import speech_recognition as sr
|
4 |
from Levenshtein import ratio
|
|
|
10 |
# Sample dataframe with sentences
|
11 |
data = {
|
12 |
"Sentences": [
|
13 |
+
"A stitch in time saves nine.",
|
|
|
14 |
"To be or not to be, that is the question.",
|
15 |
+
"Five cats were living in safe caves.",
|
16 |
+
"Hives give shelter to bees in large caves.",
|
17 |
+
"His decision to plant a rose was amazing.",
|
18 |
"She sells sea shells by the sea shore.",
|
19 |
+
"The colorful parrot likes rolling berries.",
|
|
|
|
|
20 |
"Time flies like an arrow; fruit flies like a banana.",
|
21 |
+
"Good things come to those who wait.",
|
22 |
+
"All human beings are born free and equal in dignity and rights."
|
23 |
]
|
24 |
}
|
25 |
df = pd.DataFrame(data)
|
26 |
+
user_scores = {}
|
27 |
|
28 |
def transcribe_audio(file_info):
|
29 |
r = sr.Recognizer()
|
30 |
with tempfile.NamedTemporaryFile(delete=True, suffix=".wav") as tmpfile:
|
31 |
+
sf.write(tmpfile.name, data=file_info[1], samplerate=44100, format='WAV')
|
32 |
tmpfile.seek(0)
|
33 |
with sr.AudioFile(tmpfile.name) as source:
|
34 |
audio_data = r.record(source)
|
|
|
40 |
except sr.RequestError as e:
|
41 |
return f"Could not request results; {e}"
|
42 |
|
43 |
+
def pronunciation_correction(name, expected_text, file_info):
|
|
|
44 |
user_spoken_text = transcribe_audio(file_info)
|
45 |
similarity = ratio(expected_text.lower(), user_spoken_text.lower())
|
46 |
+
score = float(f"{similarity:.2f}")
|
47 |
+
if name in user_scores:
|
48 |
+
user_scores[name].append(score) # Track scores for each user
|
|
|
|
|
|
|
|
|
|
|
49 |
else:
|
50 |
+
user_scores[name] = [score]
|
51 |
+
feedback = "Excellent pronunciation!" if score >= 0.9 else \
|
52 |
+
"Good pronunciation!" if score >= 0.7 else \
|
53 |
+
"Needs improvement." if score >= 0.5 else \
|
54 |
+
"Poor pronunciation, try to focus more on clarity."
|
55 |
+
return feedback, score
|
56 |
|
57 |
+
def calculate_average(name):
|
58 |
+
if name in user_scores and user_scores[name]:
|
59 |
+
filtered_scores = [score for score in user_scores[name] if score > 0] # Ignore zeros
|
60 |
+
average_score = sum(filtered_scores) / len(filtered_scores)
|
61 |
+
else:
|
62 |
+
average_score = 0
|
63 |
+
return f"π Great job, {name}! \n\nYour average score (excluding zeros) is: {average_score:.2f}. \nRemember, this score only focuses on the accuracy of individual sounds. \nKeep up the fun and enjoyment as you continue learning English!"
|
64 |
|
65 |
+
with gr.Blocks() as app:
|
66 |
+
name_input = gr.Textbox(label="Enter your name", placeholder="Type your name here...", value="")
|
67 |
+
with gr.Row():
|
68 |
+
sentence_dropdown = gr.Dropdown(choices=df['Sentences'].tolist(), label="Select a Sentence")
|
69 |
+
selected_sentence_output = gr.Textbox(label="Selected Text", interactive=False)
|
70 |
+
audio_input = gr.Audio(label="Upload Audio File", type="numpy")
|
71 |
+
check_pronunciation_button = gr.Button("Check Pronunciation")
|
72 |
+
pronunciation_feedback = gr.Textbox(label="Pronunciation Feedback")
|
73 |
+
pronunciation_score = gr.Number(label="Pronunciation Accuracy Score: 0 (No Match) ~ 1 (Perfect)")
|
74 |
+
complete_button = gr.Button("Complete")
|
75 |
+
average_output = gr.Textbox(label="Average Score Output", visible=True)
|
|
|
76 |
|
77 |
+
sentence_dropdown.change(lambda x: x, inputs=sentence_dropdown, outputs=selected_sentence_output)
|
78 |
+
check_pronunciation_button.click(
|
79 |
+
pronunciation_correction,
|
80 |
+
inputs=[name_input, sentence_dropdown, audio_input],
|
81 |
+
outputs=[pronunciation_feedback, pronunciation_score]
|
82 |
+
)
|
83 |
+
complete_button.click(
|
84 |
+
calculate_average,
|
85 |
+
inputs=[name_input],
|
86 |
+
outputs=average_output
|
87 |
+
)
|
88 |
|
89 |
+
app.launch(debug=True)
|