app.py

#Importing the required libraries
import gradio as gr   
import librosa 
from transformers import AutoFeatureExtractor, pipeline 

#Loading and fixing the audio file
def load_and_fix_data(input_file, model_sampling_rate):
    speech, sample_rate = librosa.load(input_file) 
    if len(speech.shape) > 1:
        speech = speech[:, 0] + speech[:, 1] 
    if sample_rate != model_sampling_rate:
        speech = librosa.resample(speech, sample_rate, model_sampling_rate)
    return speech

#Loading the feature extractor and setting up the pipeline
model_asr = "jonatasgrosman/wav2vec2-xls-r-1b-spanish"
feature_extractor = AutoFeatureExtractor.from_pretrained(model_asr)
sampling_rate = feature_extractor.sampling_rate
asr = pipeline("automatic-speech-recognition", model=model_asr)

#Instantiating the pipeline for sentiment analysis
model_sentiment_classifier = "finiteautomata/beto-sentiment-analysis"
classifier = pipeline("sentiment-analysis", model = model_sentiment_classifier)

#Defining a function for speech-to_text conversion
def speech_to_text(speech):
    audio_transcription = asr(speech, chunk_length_s = 12, stride_length_s=1)["text"]
    return audio_transcription

#Defining a function to classify sentiment of the resulting audio transcription    
def sentiment_classifier(text):
    detected_sentiment = classifier(text)[0]["label"]
    return detected_sentiment

new_line = "\n\n\n"

#Defining a function that outputs audio transcription and the result of sentiment detection module
def asr_and_sentiment_detection(input_file):
    speech = load_and_fix_data(input_file, sampling_rate)
    transcription = speech_to_text(speech)
    sentiment = sentiment_classifier(transcription)
    return f"Audio Transcription :{transcription} {new_line} Detected Sentiment: {sentiment}"


inputs = [gr.inputs.Audio(source="microphone", type="filepath", label="Record your audio")]
outputs = [gr.outputs.Textbox(label="Predicción")]
examples = [["audio_test.wav"], ["sample_audio.wav"], ["test1.wav"], ["test2.wav"], ["Example1.wav"]]
title = "Spanish ASR and Sentiment Classifier"

description = """ This is a Gradio demo for Spanish ASR and Sentiment Analysis. First, we do Speech to Text conversion, and then we perform sentiment analysis on the obtained transcription of the input audio. 

**Note regarding predicted labels : NEG --> NEGATIVE, NEU --> NEUTRAL, POS --> POSITIVE**

Pre-trained model used for Spanish ASR: [jonatasgrosman/wav2vec2-xls-r-1b-spanish](https://huggingface.co/jonatasgrosman/wav2vec2-xls-r-1b-spanish)

Pre-trained model used for Sentiment Analysis of transcribed audio: [finiteautomata/beto-sentiment-analysis](https://huggingface.co/finiteautomata/beto-sentiment-analysis)
"""


gr.Interface(
    asr_and_sentiment_detection,
    inputs = inputs,
    outputs=outputs,
    examples=examples,
    title=title,
    description=description,
    layout="horizontal",
    theme="huggingface",
).launch(enable_queue=True)