File size: 8,579 Bytes
30ca17d
 
 
 
 
 
 
 
 
 
ea87c49
30ca17d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea87c49
30ca17d
 
 
 
 
 
 
ea87c49
30ca17d
 
 
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
from transformers import T5ForConditionalGeneration, T5Tokenizer, AutoModel, AutoTokenizer
import torch
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np
import gradio as gr
from collections import Counter
import pandas as pd

# Load paraphrase model and tokenizer
model = T5ForConditionalGeneration.from_pretrained('ramsrigouthamg/t5_paraphraser')
tokenizer = T5Tokenizer.from_pretrained('t5-base', legacy=False)  # Explicitly set legacy=False

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

# Load Sentence-BERT model for semantic similarity calculation
embed_model = AutoModel.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
embed_tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
embed_model = embed_model.to(device)

# Function to get sentence embeddings
def get_sentence_embedding(sentence):
    inputs = embed_tokenizer(sentence, return_tensors="pt", padding=True).to(device)
    with torch.no_grad():
        embeddings = embed_model(**inputs).last_hidden_state.mean(dim=1)
    return embeddings

# Paraphrasing function
def paraphrase_sentence(sentence):
    if not sentence.strip():
        return "Please enter a valid sentence."
    
    # Updated prompt for statement-like output
    text = "rephrase as a statement: " + sentence
    encoding = tokenizer.encode_plus(text, padding=False, return_tensors="pt")
    input_ids, attention_masks = encoding["input_ids"].to(device), encoding["attention_mask"].to(device)

    beam_outputs = model.generate(
        input_ids=input_ids, 
        attention_mask=attention_masks,
        do_sample=True,
        max_length=128,          
        top_k=40,                
        top_p=0.85,              
        early_stopping=True,
        num_return_sequences=5   
    )

    # Decode and format paraphrases with numbering
    paraphrases = []
    for i, line in enumerate(beam_outputs, 1):
        paraphrase = tokenizer.decode(line, skip_special_tokens=True, clean_up_tokenization_spaces=True)
        paraphrases.append(f"{i}. {paraphrase}")

    return "\n".join(paraphrases)  

# Precision, Recall, and Overall Accuracy Calculation
def calculate_precision_recall_accuracy(sentences):
    total_similarity = 0
    paraphrase_count = 0
    total_precision = 0
    total_recall = 0
    
    for sentence in sentences:
        paraphrases = paraphrase_sentence(sentence).split("\n")
        
        # Get the original embedding and token counts
        original_embedding = get_sentence_embedding(sentence)
        original_tokens = Counter(sentence.lower().split())
        
        for paraphrase in paraphrases:
            if not paraphrase.strip():
                continue
            # Remove numbering before evaluation
            paraphrase_text = paraphrase.split(". ", 1)[1] if ". " in paraphrase else paraphrase
            paraphrase_embedding = get_sentence_embedding(paraphrase_text)
            similarity = cosine_similarity(original_embedding.cpu(), paraphrase_embedding.cpu())[0][0]
            total_similarity += similarity

            # Calculate precision and recall based on token overlap
            paraphrase_tokens = Counter(paraphrase_text.lower().split())
            overlap = sum((paraphrase_tokens & original_tokens).values())
            precision = overlap / sum(paraphrase_tokens.values()) if paraphrase_tokens else 0
            recall = overlap / sum(original_tokens.values()) if original_tokens else 0

            total_precision += precision
            total_recall += recall
            paraphrase_count += 1

    # Calculate averages for accuracy, precision, and recall
    overall_accuracy = (total_similarity / paraphrase_count) * 100 if paraphrase_count else 0
    avg_precision = (total_precision / paraphrase_count) * 100 if paraphrase_count else 0
    avg_recall = (total_recall / paraphrase_count) * 100 if paraphrase_count else 0

    return (f"**Overall Model Accuracy (Semantic Similarity):** {overall_accuracy:.2f}%\n"
            f"**Average Precision (Token Overlap):** {avg_precision:.2f}%\n"
            f"**Average Recall (Token Overlap):** {avg_recall:.2f}%")

# Custom CSS for aesthetic design
custom_css = """
body {
    background: linear-gradient(135deg, #e0e7ff, #c3dafe, #e0e7ff);
    font-family: 'Inter', sans-serif;
}
.gradio-container {
    max-width: 800px !important;
    margin: auto;
    padding: 20px;
    background: white;
    border-radius: 20px;
    box-shadow: 0 10px 30px rgba(0, 0, 0, 0.1);
}
h1 {
    font-size: 2.5rem;
    font-weight: 700;
    background: linear-gradient(to right, #4f46e5, #7c3aed);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    text-align: center;
    margin-bottom: 1rem;
}
textarea, input {
    border: 2px solid #e0e7ff !important;
    border-radius: 10px !important;
    padding: 15px !important;
    transition: all 0.3s ease !important;
}
textarea:hover, input:hover {
    border-color: #a5b4fc !important;
    box-shadow: 0 0 10px rgba(79, 70, 229, 0.2) !important;
}
textarea:focus, input:focus {
    border-color: #4f46e5 !important;
    box-shadow: 0 0 15px rgba(79, 70, 229, 0.3) !important;
}
button {
    background: linear-gradient(to right, #4f46e5, #7c3aed) !important;
    color: white !important;
    font-weight: 600 !important;
    padding: 12px 24px !important;
    border-radius: 10px !important;
    border: none !important;
    transition: all 0.3s ease !important;
}
button:hover {
    background: linear-gradient(to right, #4338ca, #6d28d9) !important;
    transform: scale(1.05) !important;
    box-shadow: 0 5px 15px rgba(79, 70, 229, 0.4) !important;
}
button:disabled {
    background: linear-gradient(to right, #a3a3a3, #d1d5db) !important;
    transform: none !important;
    box-shadow: none !important;
}
.output-text {
    background: #f9fafb !important;
    border-radius: 10px !important;
    padding: 15px !important;
    border: 1px solid #e5e7eb !important;
    transition: all 0.3s ease !important;
}
.output-text:hover {
    background: #eff6ff !important;
    border-color: #a5b4fc !important;
}
footer {
    display: none !important;
}
"""

# Custom JavaScript for additional interactivity
custom_js = """
<script>
document.addEventListener('DOMContentLoaded', () => {
    const textarea = document.querySelector('textarea');
    const button = document.querySelector('button');
    
    // Add typing animation effect
    textarea.addEventListener('input', () => {
        textarea.style.transform = 'scale(1.02)';
        setTimeout(() => {
            textarea.style.transform = 'scale(1)';
        }, 200);
    });
    
    // Button click animation
    button.addEventListener('click', () => {
        if (!button.disabled) {
            button.style.transform = 'scale(0.95)';
            setTimeout(() => {
                button.style.transform = 'scale(1)';
            }, 200);
        }
    });
});
</script>
"""

# Define Gradio UI with enhanced aesthetics
with gr.Blocks(theme=gr.themes.Soft(), css=custom_css, js=custom_js) as demo:
    gr.Markdown(
        """
        # PARA-GEN: Aesthetic Paraphraser
        Enter a sentence below to generate five beautifully rephrased statements.
        """
    )
    
    with gr.Row():
        with gr.Column(scale=3):
            input_text = gr.Textbox(
                label="Input Sentence",
                placeholder="Type your sentence here...",
                lines=4,
                max_lines=4
            )
            paraphrase_button = gr.Button("Generate Paraphrases")
        
        with gr.Column(scale=2):
            output_text = gr.Textbox(
                label="Paraphrased Results",
                lines=10,
                interactive=False
            )
    
    with gr.Accordion("Model Performance Metrics", open=False):
        metrics_output = gr.Markdown()
    
    # Define button click behavior
    paraphrase_button.click(
        fn=paraphrase_sentence,
        inputs=input_text,
        outputs=output_text
    )
    
    # Calculate and display metrics on load without _js
    test_sentences = [
        "The quick brown fox jumps over the lazy dog.",
        "Artificial intelligence is transforming industries.",
        "The weather is sunny and warm today.",
        "He enjoys reading books on machine learning.",
        "The stock market fluctuates daily due to various factors."
    ]
    metrics_output.value = calculate_precision_recall_accuracy(test_sentences)

# Launch Gradio app
demo.launch(share=False)