Spaces:
Runtime error
Runtime error
Commit
•
e2e3b5f
1
Parent(s):
729ba94
added application file
Browse files
app.py
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
|
3 |
+
import crepe
|
4 |
+
import spacy
|
5 |
+
import librosa
|
6 |
+
import gradio as gr
|
7 |
+
import pandas as pd
|
8 |
+
from transformers import pipeline
|
9 |
+
|
10 |
+
asr = pipeline('automatic-speech-recognition', model='facebook/wav2vec2-large-960h-lv60-self')
|
11 |
+
emo = pipeline('sentiment-analysis', model='arpanghoshal/EmoRoBERTa')
|
12 |
+
lang_model = spacy.load("en_core_web_sm")
|
13 |
+
|
14 |
+
def transcribe_and_describe(audio):
|
15 |
+
|
16 |
+
audio, sr = librosa.load(audio, sr=16000)
|
17 |
+
|
18 |
+
text = asr(audio)['text']
|
19 |
+
|
20 |
+
doc = lang_model(text)
|
21 |
+
filler_words = [token.text for token in doc if token.pos_ == 'INTJ']
|
22 |
+
filler_word_pr = len(filler_words) / len(doc)
|
23 |
+
|
24 |
+
flatness = pd.DataFrame(librosa.feature.spectral_flatness(y=audio).T).describe().T
|
25 |
+
loudness = pd.DataFrame(librosa.feature.rms(audio).T).describe().T
|
26 |
+
time, frequency, confidence, activation = crepe.predict(audio, sr)
|
27 |
+
frequency = pd.DataFrame(frequency.T).describe().T
|
28 |
+
|
29 |
+
mean_spectral_flatness = flatness.loc[0, 'mean']
|
30 |
+
spectral_flatness_std = flatness.loc[0, 'std']
|
31 |
+
mean_pitch = frequency.loc[0, 'mean']
|
32 |
+
pitch_std = frequency.loc[0, 'std']
|
33 |
+
mean_volume = loudness.loc[0, 'mean']
|
34 |
+
volume_std = loudness.loc[0, 'std']
|
35 |
+
|
36 |
+
words_per_minute = len(text.split(" ")) / (librosa.get_duration(audio, sr) / 60)
|
37 |
+
|
38 |
+
emotion = emo(text)[0]['label']
|
39 |
+
|
40 |
+
return (text, filler_word_pr, words_per_minute, mean_pitch, pitch_std, mean_volume, volume_std, mean_spectral_flatness, spectral_flatness_std, emotion)
|
41 |
+
|
42 |
+
gr.Interface(
|
43 |
+
fn=transcribe_and_describe,
|
44 |
+
inputs=gr.Audio(source="microphone", type="filepath"),
|
45 |
+
outputs=[
|
46 |
+
gr.Text(label="Transcription"),
|
47 |
+
gr.Text(label="Rate of Speech (WPM)"),
|
48 |
+
gr.Text(label="Filler Word Percent"),
|
49 |
+
gr.Text(label="Mean Pitch (Hz)"),
|
50 |
+
gr.Text(label="Pitch Variation (Hz)"),
|
51 |
+
gr.Text(label="Mean Volume (W)"),
|
52 |
+
gr.Text(label="Volume Variation (W)"),
|
53 |
+
gr.Text(label="Mean Spectral Flatness (dB)"),
|
54 |
+
gr.Text(label="Spectral Flatness Variation (dB)"),
|
55 |
+
gr.Text(label="Emotion")
|
56 |
+
]
|
57 |
+
).launch()
|