File size: 10,623 Bytes
39028c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
"""
Intelligent model selection based on document characteristics
"""

import logging
from typing import Optional, Dict, Any
from enum import Enum

logger = logging.getLogger(__name__)


class ModelCategory(Enum):
    """Model categories by speed/complexity tradeoff."""
    LIGHTWEIGHT = "lightweight"
    BALANCED = "balanced"
    ADVANCED = "advanced"


class DocumentComplexity(Enum):
    """Document complexity levels."""
    SIMPLE = "simple"
    MODERATE = "moderate"
    COMPLEX = "complex"
    VERY_COMPLEX = "very_complex"


class ModelSelector:
    """Intelligent model selector based on document characteristics."""
    
    LIGHTWEIGHT_MODELS = {
        't5-small': {'params': 60e6, 'speed': 'very_fast', 'quality': 'good'},
        'distilbart-cnn-6-6': {'params': 82e6, 'speed': 'very_fast', 'quality': 'good'},
        'bart-base': {'params': 140e6, 'speed': 'fast', 'quality': 'very_good'},
    }
    
    BALANCED_MODELS = {
        't5-base': {'params': 220e6, 'speed': 'fast', 'quality': 'excellent'},
        'pegasus-arxiv': {'params': 568e6, 'speed': 'moderate', 'quality': 'excellent'},
        'bart-large-cnn': {'params': 406e6, 'speed': 'moderate', 'quality': 'excellent'},
    }
    
    ADVANCED_MODELS = {
        'llama-7b': {'params': 7e9, 'speed': 'slow', 'quality': 'outstanding'},
        'gpt-3.5': {'params': 175e9, 'speed': 'slow', 'quality': 'outstanding'},
        't5-large': {'params': 770e6, 'speed': 'moderate', 'quality': 'outstanding'},
    }
    
    COMPLEXITY_THRESHOLDS = {
        'word_count': {
            'simple': 500,
            'moderate': 2000,
            'complex': 5000,
            'very_complex': float('inf')
        },
        'sentence_length': {
            'simple': 15,
            'moderate': 25,
            'complex': 40,
            'very_complex': float('inf')
        },
        'vocabulary_complexity': {
            'simple': 0.3,
            'moderate': 0.5,
            'complex': 0.7,
            'very_complex': 1.0
        }
    }
    
    def __init__(self):
        """Initialize model selector."""
        self.current_model = None
        self.current_complexity = None
        self.use_rag = False
    
    def analyze_document(self, text: str) -> Dict[str, Any]:
        """
        Analyze document to determine complexity.
        
        Args:
            text: Document text
            
        Returns:
            Dictionary with complexity analysis
        """
        words = text.split()
        sentences = text.split('.')
        
        word_count = len(words)
        sentence_count = len([s for s in sentences if s.strip()])
        
        avg_sentence_length = word_count / max(sentence_count, 1)
        
        unique_words = len(set(w.lower() for w in words))
        vocabulary_richness = unique_words / max(word_count, 1)
        
        analysis = {
            'word_count': word_count,
            'sentence_count': sentence_count,
            'avg_sentence_length': avg_sentence_length,
            'vocabulary_richness': vocabulary_richness,
        }
        
        return analysis
    
    def determine_complexity(self, analysis: Dict[str, Any]) -> DocumentComplexity:
        """
        Determine document complexity from analysis.
        
        Args:
            analysis: Document analysis dictionary
            
        Returns:
            DocumentComplexity enum
        """
        word_count = analysis['word_count']
        avg_sentence_length = analysis['avg_sentence_length']
        vocab_richness = analysis['vocabulary_richness']
        
        complexity_score = 0
        
        if word_count > self.COMPLEXITY_THRESHOLDS['word_count']['very_complex']:
            complexity_score += 3
        elif word_count > self.COMPLEXITY_THRESHOLDS['word_count']['complex']:
            complexity_score += 2.5
        elif word_count > self.COMPLEXITY_THRESHOLDS['word_count']['moderate']:
            complexity_score += 1.5
        elif word_count > self.COMPLEXITY_THRESHOLDS['word_count']['simple']:
            complexity_score += 0.5
        
        if avg_sentence_length > self.COMPLEXITY_THRESHOLDS['sentence_length']['very_complex']:
            complexity_score += 3
        elif avg_sentence_length > self.COMPLEXITY_THRESHOLDS['sentence_length']['complex']:
            complexity_score += 2.5
        elif avg_sentence_length > self.COMPLEXITY_THRESHOLDS['sentence_length']['moderate']:
            complexity_score += 1.5
        elif avg_sentence_length > self.COMPLEXITY_THRESHOLDS['sentence_length']['simple']:
            complexity_score += 0.5
        
        if vocab_richness > self.COMPLEXITY_THRESHOLDS['vocabulary_complexity']['very_complex']:
            complexity_score += 3
        elif vocab_richness > self.COMPLEXITY_THRESHOLDS['vocabulary_complexity']['complex']:
            complexity_score += 2.5
        elif vocab_richness > self.COMPLEXITY_THRESHOLDS['vocabulary_complexity']['moderate']:
            complexity_score += 1.5
        elif vocab_richness > self.COMPLEXITY_THRESHOLDS['vocabulary_complexity']['simple']:
            complexity_score += 0.5
        
        if complexity_score < 2:
            return DocumentComplexity.SIMPLE
        elif complexity_score < 4:
            return DocumentComplexity.MODERATE
        elif complexity_score < 6:
            return DocumentComplexity.COMPLEX
        else:
            return DocumentComplexity.VERY_COMPLEX
    
    def select_model(self, text: str, prefer_fast: bool = True) -> Dict[str, Any]:
        """
        Select best model based on document.
        
        Args:
            text: Document text
            prefer_fast: If True, prefer speed over quality
            
        Returns:
            Dictionary with model recommendation and settings
        """
        analysis = self.analyze_document(text)
        complexity = self.determine_complexity(analysis)
        
        self.current_complexity = complexity
        
        logger.info(f"Document complexity: {complexity.value}")
        logger.info(f"Analysis: word_count={analysis['word_count']}, "
                   f"avg_sentence_length={analysis['avg_sentence_length']:.1f}, "
                   f"vocabulary_richness={analysis['vocabulary_richness']:.2f}")
        
        if complexity == DocumentComplexity.SIMPLE:
            return self._get_lightweight_recommendation(prefer_fast)
        elif complexity == DocumentComplexity.MODERATE:
            return self._get_balanced_recommendation(prefer_fast)
        elif complexity == DocumentComplexity.COMPLEX:
            return self._get_advanced_recommendation(prefer_fast)
        else:
            return self._get_advanced_recommendation_with_rag(prefer_fast)
    
    def _get_lightweight_recommendation(self, prefer_fast: bool) -> Dict[str, Any]:
        """Recommend lightweight model for simple documents."""
        if prefer_fast:
            model = 't5-small'
        else:
            model = 'bart-base'
        
        return {
            'model': model,
            'category': ModelCategory.LIGHTWEIGHT,
            'use_rag': False,
            'num_beams': 2,
            'max_length': 100,
            'reason': 'Simple document - lightweight model sufficient',
            'estimated_time': '1-2 seconds'
        }
    
    def _get_balanced_recommendation(self, prefer_fast: bool) -> Dict[str, Any]:
        """Recommend balanced model for moderate complexity documents."""
        if prefer_fast:
            model = 't5-base'
        else:
            model = 'bart-large-cnn'
        
        return {
            'model': model,
            'category': ModelCategory.BALANCED,
            'use_rag': False,
            'num_beams': 4,
            'max_length': 150,
            'reason': 'Moderate complexity - balanced model recommended',
            'estimated_time': '2-5 seconds'
        }
    
    def _get_advanced_recommendation(self, prefer_fast: bool) -> Dict[str, Any]:
        """Recommend advanced model for complex documents."""
        if prefer_fast:
            model = 'pegasus-arxiv'
        else:
            model = 't5-large'
        
        return {
            'model': model,
            'category': ModelCategory.ADVANCED,
            'use_rag': True,
            'num_beams': 4,
            'max_length': 200,
            'reason': 'Complex document - advanced model with RAG recommended',
            'estimated_time': '3-7 seconds'
        }
    
    def _get_advanced_recommendation_with_rag(self, prefer_fast: bool) -> Dict[str, Any]:
        """Recommend advanced model with RAG for very complex documents."""
        return {
            'model': 't5-large',
            'category': ModelCategory.ADVANCED,
            'use_rag': True,
            'num_beams': 4,
            'max_length': 250,
            'reason': 'Very complex document - RAG pipeline required for context awareness',
            'estimated_time': '5-10 seconds',
            'retrieval_chunks': 5,
            'chunk_overlap': 100
        }
    
    def should_use_rag(self, complexity: DocumentComplexity) -> bool:
        """
        Determine if RAG should be used.
        
        Args:
            complexity: Document complexity
            
        Returns:
            Boolean indicating whether to use RAG
        """
        return complexity in [DocumentComplexity.COMPLEX, DocumentComplexity.VERY_COMPLEX]
    
    def get_model_info(self, model_name: str) -> Optional[Dict[str, Any]]:
        """
        Get information about a model.
        
        Args:
            model_name: Model name
            
        Returns:
            Model information or None if not found
        """
        for models_dict in [self.LIGHTWEIGHT_MODELS, self.BALANCED_MODELS, self.ADVANCED_MODELS]:
            if model_name in models_dict:
                return models_dict[model_name]
        
        return None
    
    def recommend_settings(self, text: str, quality_preference: str = 'balanced') -> Dict[str, Any]:
        """
        Get recommended settings for document summarization.
        
        Args:
            text: Document text
            quality_preference: 'speed', 'balanced', or 'quality'
            
        Returns:
            Recommended settings dictionary
        """
        prefer_fast = quality_preference == 'speed'
        recommendation = self.select_model(text, prefer_fast)
        
        model_info = self.get_model_info(recommendation['model'])
        if model_info:
            recommendation['model_info'] = model_info
        
        return recommendation