scripts/readability_scorer.py
10.6 KB| 1 | #!/usr/bin/env python3 |
| 2 | """ |
| 3 | readability_scorer.py — Readability metrics for marketing copy |
| 4 | Usage: |
| 5 | python3 readability_scorer.py --file copy.txt |
| 6 | echo "Your text here" | python3 readability_scorer.py |
| 7 | python3 readability_scorer.py # demo mode |
| 8 | python3 readability_scorer.py --json |
| 9 | """ |
| 10 | |
| 11 | import argparse |
| 12 | import json |
| 13 | import math |
| 14 | import re |
| 15 | import sys |
| 16 | |
| 17 | |
| 18 | # --------------------------------------------------------------------------- |
| 19 | # Word lists |
| 20 | # --------------------------------------------------------------------------- |
| 21 | |
| 22 | FILLER_WORDS = [ |
| 23 | "very", "really", "just", "actually", "basically", "literally", |
| 24 | "honestly", "totally", "absolutely", "definitely", "certainly", |
| 25 | "obviously", "clearly", "quite", "rather", "somewhat", "fairly", |
| 26 | "pretty", "simply", "truly", "genuinely", "essentially", |
| 27 | ] |
| 28 | |
| 29 | # Simple passive voice detection: "was/were/is/are/been/being + past participle" |
| 30 | PASSIVE_PATTERN = re.compile( |
| 31 | r"\b(was|were|is|are|been|being|be|am)\s+(\w+ed|known|written|built|made|done|seen|given|taken|brought|thought|found|put|set|cut|read|let|hit|hurt|cost|led|felt|kept|left|meant|sent|spent|stood|told|wore|won|beat|lost|broke|chose|drove|flew|froze|grew|hid|rang|rode|rose|ran|sank|sang|spoke|swore|swam|threw|woke|wrote)\b", |
| 32 | re.IGNORECASE, |
| 33 | ) |
| 34 | |
| 35 | ADVERB_PATTERN = re.compile(r"\b\w+ly\b", re.IGNORECASE) |
| 36 | |
| 37 | # Syllable estimation: count vowel groups |
| 38 | def count_syllables(word: str) -> int: |
| 39 | word = word.lower().strip(".,!?;:\"'") |
| 40 | if not word: |
| 41 | return 0 |
| 42 | # Silent e |
| 43 | if word.endswith("e") and len(word) > 2: |
| 44 | word = word[:-1] |
| 45 | count = len(re.findall(r"[aeiou]+", word)) |
| 46 | return max(1, count) |
| 47 | |
| 48 | |
| 49 | def split_sentences(text: str) -> list: |
| 50 | # Split on sentence-ending punctuation |
| 51 | parts = re.split(r"(?<=[.!?])\s+", text.strip()) |
| 52 | return [p.strip() for p in parts if p.strip()] |
| 53 | |
| 54 | |
| 55 | def split_words(text: str) -> list: |
| 56 | return re.findall(r"\b[a-zA-Z]+\b", text) |
| 57 | |
| 58 | |
| 59 | # --------------------------------------------------------------------------- |
| 60 | # Metrics |
| 61 | # --------------------------------------------------------------------------- |
| 62 | |
| 63 | def flesch_reading_ease(avg_sentence_len: float, avg_syllables: float) -> float: |
| 64 | """Flesch Reading Ease formula.""" |
| 65 | score = 206.835 - (1.015 * avg_sentence_len) - (84.6 * avg_syllables) |
| 66 | return round(max(0.0, min(100.0, score)), 1) |
| 67 | |
| 68 | |
| 69 | def flesch_kincaid_grade(avg_sentence_len: float, avg_syllables: float) -> float: |
| 70 | """Flesch-Kincaid Grade Level formula.""" |
| 71 | grade = (0.39 * avg_sentence_len) + (11.8 * avg_syllables) - 15.59 |
| 72 | return round(max(0.0, grade), 1) |
| 73 | |
| 74 | |
| 75 | def ease_label(score: float) -> str: |
| 76 | if score >= 90: return "Very Easy (5th grade)" |
| 77 | if score >= 80: return "Easy (6th grade)" |
| 78 | if score >= 70: return "Fairly Easy (7th grade)" |
| 79 | if score >= 60: return "Standard (8-9th grade)" |
| 80 | if score >= 50: return "Fairly Difficult (10-12th grade)" |
| 81 | if score >= 30: return "Difficult (College)" |
| 82 | return "Very Confusing (Professional)" |
| 83 | |
| 84 | |
| 85 | def analyze_text(text: str) -> dict: |
| 86 | sentences = split_sentences(text) |
| 87 | words = split_words(text) |
| 88 | |
| 89 | if not words: |
| 90 | return {"error": "No readable text found."} |
| 91 | |
| 92 | num_sentences = max(1, len(sentences)) |
| 93 | num_words = len(words) |
| 94 | |
| 95 | # Syllables |
| 96 | syllable_counts = [count_syllables(w) for w in words] |
| 97 | total_syllables = sum(syllable_counts) |
| 98 | |
| 99 | avg_sentence_len = num_words / num_sentences |
| 100 | avg_word_len = sum(len(w) for w in words) / num_words |
| 101 | avg_syllables_per_word = total_syllables / num_words |
| 102 | |
| 103 | fre = flesch_reading_ease(avg_sentence_len, avg_syllables_per_word) |
| 104 | fk_grade = flesch_kincaid_grade(avg_sentence_len, avg_syllables_per_word) |
| 105 | |
| 106 | # Passive voice |
| 107 | passive_matches = PASSIVE_PATTERN.findall(text) |
| 108 | passive_count = len(passive_matches) |
| 109 | passive_pct = round(passive_count / num_sentences * 100, 1) |
| 110 | |
| 111 | # Adverbs |
| 112 | adverb_matches = ADVERB_PATTERN.findall(text) |
| 113 | # Filter obvious non-adverbs |
| 114 | non_adverb = {"family", "early", "only", "likely", "nearly", "really", |
| 115 | "daily", "weekly", "monthly", "yearly", "friendly", "lovely", |
| 116 | "lonely", "lively", "elderly", "costly"} |
| 117 | adverbs = [a for a in adverb_matches if a.lower() not in non_adverb] |
| 118 | adverb_density = round(len(adverbs) / num_words * 100, 1) |
| 119 | |
| 120 | # Filler words |
| 121 | text_lower = text.lower() |
| 122 | word_tokens_lower = [w.lower() for w in words] |
| 123 | filler_found = {fw: word_tokens_lower.count(fw) for fw in FILLER_WORDS if fw in word_tokens_lower} |
| 124 | filler_total = sum(filler_found.values()) |
| 125 | |
| 126 | # Scoring: |
| 127 | # FRE already 0-100 (higher = easier = better for marketing copy) |
| 128 | # Target for marketing: 60-80 range |
| 129 | fre_score = fre # use as-is |
| 130 | |
| 131 | return { |
| 132 | "stats": { |
| 133 | "word_count": num_words, |
| 134 | "sentence_count": num_sentences, |
| 135 | "avg_sentence_length": round(avg_sentence_len, 1), |
| 136 | "avg_word_length": round(avg_word_len, 1), |
| 137 | "avg_syllables_per_word": round(avg_syllables_per_word, 2), |
| 138 | }, |
| 139 | "flesch_reading_ease": { |
| 140 | "score": fre, |
| 141 | "label": ease_label(fre), |
| 142 | "target": "60-80 for most marketing copy", |
| 143 | }, |
| 144 | "flesch_kincaid_grade": { |
| 145 | "grade_level": fk_grade, |
| 146 | "note": f"Equivalent to grade {fk_grade} reading level", |
| 147 | }, |
| 148 | "passive_voice": { |
| 149 | "count": passive_count, |
| 150 | "percentage": passive_pct, |
| 151 | "target": "<10%", |
| 152 | "pass": passive_pct < 10, |
| 153 | }, |
| 154 | "adverb_density": { |
| 155 | "count": len(adverbs), |
| 156 | "percentage": adverb_density, |
| 157 | "examples": list(set(adverbs))[:8], |
| 158 | "target": "<5%", |
| 159 | "pass": adverb_density < 5, |
| 160 | }, |
| 161 | "filler_words": { |
| 162 | "total_count": filler_total, |
| 163 | "breakdown": filler_found, |
| 164 | "target": "0-3 per 100 words", |
| 165 | "per_100_words": round(filler_total / num_words * 100, 1), |
| 166 | }, |
| 167 | "overall_score": round(fre), |
| 168 | } |
| 169 | |
| 170 | |
| 171 | # --------------------------------------------------------------------------- |
| 172 | # Demo text |
| 173 | # --------------------------------------------------------------------------- |
| 174 | |
| 175 | DEMO_TEXT = """ |
| 176 | Marketing copy needs to be clear, direct, and persuasive. When you write for your audience, |
| 177 | you should always think about what they actually want to hear. Really good copy is basically |
| 178 | about solving problems. It is very important to avoid using overly complicated language that |
| 179 | might confuse the reader. |
| 180 | |
| 181 | The best headlines are written by experts who truly understand their customers. A strong |
| 182 | call-to-action is absolutely essential for any landing page. You need to make sure that |
| 183 | every single word is earning its place on the page. |
| 184 | |
| 185 | Studies show that shorter sentences improve comprehension. The average reader processes |
| 186 | information faster when sentences contain fewer than 20 words. This is genuinely proven |
| 187 | by research. Passive voice constructions are often used by writers who want to sound |
| 188 | authoritative, but they can actually make copy feel distant and unclear. |
| 189 | |
| 190 | Focus on benefits, not features. Tell the reader what they will gain. Use numbers when |
| 191 | you can — "save 3 hours per week" beats "save time" every single time. Specificity |
| 192 | builds trust. Vague promises are ignored. |
| 193 | """ |
| 194 | |
| 195 | |
| 196 | # --------------------------------------------------------------------------- |
| 197 | # Main |
| 198 | # --------------------------------------------------------------------------- |
| 199 | |
| 200 | def main(): |
| 201 | parser = argparse.ArgumentParser( |
| 202 | description="Readability scorer for marketing copy — Flesch, passive voice, filler words." |
| 203 | ) |
| 204 | parser.add_argument("--file", help="Path to text file") |
| 205 | parser.add_argument("--json", action="store_true", help="Output as JSON") |
| 206 | args = parser.parse_args() |
| 207 | |
| 208 | if args.file: |
| 209 | with open(args.file, "r", encoding="utf-8", errors="replace") as f: |
| 210 | text = f.read() |
| 211 | elif not sys.stdin.isatty(): |
| 212 | text = sys.stdin.read() |
| 213 | if not text.strip(): |
| 214 | text = DEMO_TEXT |
| 215 | if not args.json: |
| 216 | print("No input provided — running in demo mode.\n") |
| 217 | else: |
| 218 | text = DEMO_TEXT |
| 219 | if not args.json: |
| 220 | print("No input provided — running in demo mode.\n") |
| 221 | |
| 222 | result = analyze_text(text) |
| 223 | |
| 224 | if "error" in result: |
| 225 | print(f"Error: {result['error']}", file=sys.stderr) |
| 226 | sys.exit(1) |
| 227 | |
| 228 | if args.json: |
| 229 | print(json.dumps(result, indent=2)) |
| 230 | return |
| 231 | |
| 232 | fre = result["flesch_reading_ease"] |
| 233 | fk = result["flesch_kincaid_grade"] |
| 234 | stats = result["stats"] |
| 235 | passive = result["passive_voice"] |
| 236 | adverbs = result["adverb_density"] |
| 237 | fillers = result["filler_words"] |
| 238 | score = result["overall_score"] |
| 239 | |
| 240 | PASS = "✅" |
| 241 | FAIL = "❌" |
| 242 | |
| 243 | print("=" * 62) |
| 244 | print(f" READABILITY REPORT Flesch Score: {fre['score']}/100") |
| 245 | print("=" * 62) |
| 246 | print(f" {fre['label']}") |
| 247 | print(f" Target: {fre['target']}") |
| 248 | print() |
| 249 | print(f" 📊 Stats") |
| 250 | print(f" Words: {stats['word_count']}") |
| 251 | print(f" Sentences: {stats['sentence_count']}") |
| 252 | print(f" Avg sentence length:{stats['avg_sentence_length']} words") |
| 253 | print(f" Avg word length: {stats['avg_word_length']} chars") |
| 254 | print(f" Syllables/word: {stats['avg_syllables_per_word']}") |
| 255 | print() |
| 256 | print(f" 📐 Flesch-Kincaid Grade Level: {fk['grade_level']}") |
| 257 | print(f" {fk['note']}") |
| 258 | print() |
| 259 | |
| 260 | pv_icon = PASS if passive["pass"] else FAIL |
| 261 | print(f" {pv_icon} Passive Voice: {passive['count']} instances ({passive['percentage']}%)") |
| 262 | print(f" Target: {passive['target']}") |
| 263 | |
| 264 | av_icon = PASS if adverbs["pass"] else FAIL |
| 265 | print(f" {av_icon} Adverb Density: {adverbs['count']} adverbs ({adverbs['percentage']}%)") |
| 266 | if adverbs["examples"]: |
| 267 | print(f" Examples: {', '.join(adverbs['examples'][:5])}") |
| 268 | |
| 269 | filler_ok = fillers["per_100_words"] <= 3 |
| 270 | fw_icon = PASS if filler_ok else FAIL |
| 271 | print(f" {fw_icon} Filler Words: {fillers['total_count']} total ({fillers['per_100_words']} per 100 words)") |
| 272 | if fillers["breakdown"]: |
| 273 | top = sorted(fillers["breakdown"].items(), key=lambda x: -x[1])[:5] |
| 274 | print(f" Top: {', '.join(f'{w}({c})' for w,c in top)}") |
| 275 | |
| 276 | print() |
| 277 | print("=" * 62) |
| 278 | score_bar_len = round(score / 10) |
| 279 | bar = "█" * score_bar_len + "░" * (10 - score_bar_len) |
| 280 | print(f" Readability Score: [{bar}] {score}/100") |
| 281 | print("=" * 62) |
| 282 | |
| 283 | |
| 284 | if __name__ == "__main__": |
| 285 | main() |
文件内容是平台检查时读出的文本，只供查看。平台不会运行包里的任何脚本或程序。
