File size: 5,731 Bytes
5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 53f1073 5126882 e089e5b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 |
import os
import json
import time
import numpy as np
import sounddevice as sd
from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
from huggingface_hub import login
from product_recommender import ProductRecommender
from objection_handler import load_objections, check_objections
from objection_handler import ObjectionHandler
from env_setup import config
from sentence_transformers import SentenceTransformer
from scipy.io.wavfile import write
from dotenv import load_dotenv
# Load environment variables
load_dotenv()
# Hugging Face API setup
huggingface_api_key = config["huggingface_api_key"]
login(token=huggingface_api_key)
# Sentiment Analysis Model
model_name = "tabularisai/multilingual-sentiment-analysis"
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
sentiment_analyzer = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
# Function to analyze sentiment
def preprocess_text(text):
"""Preprocess text for better sentiment analysis."""
return text.strip().lower()
def analyze_sentiment(text):
"""Analyze sentiment of the text using Hugging Face model."""
try:
if not text.strip():
return "NEUTRAL", 0.0
processed_text = preprocess_text(text)
result = sentiment_analyzer(processed_text)[0]
print(f"Sentiment Analysis Result: {result}")
# Map raw labels to sentiments
sentiment_map = {
'Very Negative': "NEGATIVE",
'Negative': "NEGATIVE",
'Neutral': "NEUTRAL",
'Positive': "POSITIVE",
'Very Positive': "POSITIVE"
}
sentiment = sentiment_map.get(result['label'], "NEUTRAL")
return sentiment, result['score']
except Exception as e:
print(f"Error in sentiment analysis: {e}")
return "NEUTRAL", 0.5
def record_audio(duration=5, sample_rate=44100):
"""Record audio for a specified duration."""
print("Recording audio...")
audio = sd.rec(int(duration * sample_rate), samplerate=sample_rate, channels=1, dtype='float32')
sd.wait() # Wait for the recording to finish
print("Recording completed.")
return np.squeeze(audio)
def transcribe_audio(audio, sample_rate=44100):
"""Transcribe recorded audio using a speech-to-text API."""
try:
# Save audio to a temporary file for transcription
audio_file = "temp_audio.wav"
write(audio_file, sample_rate, audio)
# Call external transcription service (e.g., Whisper, AssemblyAI, or Google)
transcription = "Example transcription text from audio." # Placeholder
return transcription
except Exception as e:
print(f"Error in audio transcription: {e}")
return ""
def transcribe_with_chunks(objections_dict):
"""Perform real-time transcription with sentiment analysis."""
print("Say 'start listening' to begin transcription. Say 'stop listening' to stop.")
is_listening = False
chunks = []
current_chunk = []
chunk_start_time = time.time()
# Initialize handlers with semantic search capabilities
objection_handler = ObjectionHandler("path_to_objections.csv")
product_recommender = ProductRecommender("path_to_recommendations.csv")
# Load the embeddings model once
model = SentenceTransformer('all-MiniLM-L6-v2')
try:
while True:
if not is_listening:
command = input("Enter 'start' to begin listening or 'stop' to quit: ").lower()
if command == "start":
is_listening = True
print("Listening started. Speak into the microphone.")
continue
elif command == "stop":
break
# Record and process audio in chunks
audio_data = record_audio(duration=5)
text = transcribe_audio(audio_data)
if text.strip():
print(f"Transcription: {text}")
current_chunk.append(text)
if time.time() - chunk_start_time > 3:
if current_chunk:
chunk_text = " ".join(current_chunk)
# Process sentiment
sentiment, score = analyze_sentiment(chunk_text)
chunks.append((chunk_text, sentiment, score))
# Handle objections and recommendations
query_embedding = model.encode([chunk_text])
responses = objection_handler.handle_objection(chunk_text)
recommendations = product_recommender.get_recommendations(chunk_text)
# Print results
if responses:
print("\nSuggested Response:")
for response in responses:
print(f"→ {response}")
if recommendations:
print("\nRecommendations for this response:")
for idx, rec in enumerate(recommendations, 1):
print(f"{idx}. {rec}")
print("\n")
current_chunk = []
chunk_start_time = time.time()
except KeyboardInterrupt:
print("\nExiting...")
return chunks
if __name__ == "__main__":
objections_file_path = "path_to_objections.csv"
objections_dict = load_objections(objections_file_path)
transcribed_chunks = transcribe_with_chunks(objections_dict)
print("Final transcriptions and sentiments:", transcribed_chunks)
|