File size: 7,664 Bytes
5126882
 
 
3470239
5126882
 
 
b4e4003
5126882
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3470239
 
 
5126882
 
 
 
 
 
 
 
 
 
3470239
5126882
 
3470239
5126882
3470239
5126882
 
 
 
 
 
 
 
3470239
5126882
 
3470239
5126882
 
 
 
 
70d9fc1
b4e4003
 
 
 
70d9fc1
 
 
 
 
 
 
 
b4e4003
 
 
 
 
 
 
 
 
 
 
70d9fc1
 
 
 
 
 
 
 
 
 
 
 
 
 
5126882
70d9fc1
3470239
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5126882
 
 
3470239
5126882
 
b4e4003
5126882
 
3470239
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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
import os
import json
import time
from speech_recognition import Recognizer, Microphone, AudioData, UnknownValueError, RequestError
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 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)

# Speech Recognition Setup
recognizer = Recognizer()

# 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 transcribe_with_chunks(objections_dict):
    print("Note: If microphone access fails, please use alternative input.")
    chunks = []
    current_chunk = []
    chunk_start_time = time.time()
    is_listening = False

    try:
        # Try to list available microphones
        available_mics = Microphone.list_microphone_names()
        print(f"Available microphones: {available_mics}")
    except Exception as e:
        print(f"Could not detect microphones: {e}")

    # Replace hardcoded path with environment variable or relative path
    objection_file_path = config.get("OBJECTION_DATA_PATH", "objections.csv")
    product_file_path = config.get("PRODUCT_DATA_PATH", "recommendations.csv")

    # Initialize handlers with semantic search capabilities
    objection_handler = ObjectionHandler(objection_file_path)
    product_recommender = ProductRecommender(product_file_path)

    # Load the embeddings model once
    model = SentenceTransformer('all-MiniLM-L6-v2')

    try:
        # Try multiple device indices
        mic = None
        for device_index in range(10):  # Try first 10 device indices
            try:
                mic = Microphone(device_index=device_index)
                print(f"Using microphone at device index {device_index}")
                break
            except Exception:
                continue

        if mic is None:
            print("No microphone available. Please provide text input.")
            return []

        with mic as source:
            recognizer.adjust_for_ambient_noise(source)
            print("Microphone calibrated. Please speak.")
            
            while True:
                print("Listening for speech...")
                try:
                    audio_data = recognizer.listen(source, timeout=5)
                    text = recognizer.recognize_google(audio_data)

                    if "start listening" in text.lower():
                        is_listening = True
                        print("Listening started. Speak into the microphone.")
                        continue
                    elif "stop listening" in text.lower():
                        is_listening = False
                        print("Listening stopped.")
                        if current_chunk:
                            chunk_text = " ".join(current_chunk)
                            sentiment, score = analyze_sentiment(chunk_text)
                            chunks.append((chunk_text, sentiment, score))
                            current_chunk = []
                        continue

                    if is_listening and 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)
                                
                                # Always process sentiment
                                sentiment, score = analyze_sentiment(chunk_text)
                                chunks.append((chunk_text, sentiment, score))

                                # Get objection responses and check similarity score
                                query_embedding = model.encode([chunk_text])
                                distances, indices = objection_handler.index.search(query_embedding, 1)
                                
                                # If similarity is high enough, show objection response
                                if distances[0][0] < 1.5:  # Threshold for similarity
                                    responses = objection_handler.handle_objection(chunk_text)
                                    if responses:
                                        print("\nSuggested Response:")
                                        for response in responses:
                                            print(f"→ {response}")
                                
                                # Get product recommendations and check similarity score
                                distances, indices = product_recommender.index.search(query_embedding, 1)
                                
                                # If similarity is high enough, show recommendations
                                if distances[0][0] < 1.5:  # Threshold for similarity
                                    recommendations = product_recommender.get_recommendations(chunk_text)
                                    if recommendations:
                                        print(f"\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 UnknownValueError:
                    print("Could not understand the audio.")
                except RequestError as e:
                    print(f"Could not request results from Google Speech Recognition service; {e}")

    except KeyboardInterrupt:
        print("\nExiting...")
        return chunks

if __name__ == "__main__":
    objections_file_path = config.get("OBJECTION_DATA_PATH", "objections.csv")
    objections_dict = load_objections(objections_file_path)
    transcribed_chunks = transcribe_with_chunks(objections_dict)
    print("Final transcriptions and sentiments:", transcribed_chunks)