How to Build Smarter Multilingual Text Wrapping with BudouX Through Parsing, HTML Rendering, Model Introspection, and Toy Training

import subprocess, sys
def pip(*pkgs):
   subprocess.check_call([sys.executable, "-m", "pip", "install", "-q", *pkgs])
pip("budoux")


import json, time, textwrap, html, random, re, os, tempfile
from pathlib import Path
import budoux
from IPython.show import HTML, show, Markdown


print(f" BudouX model: {budoux.__version__ if hasattr(budoux,'__version__') else 'put in'}")


def header(title):
   show(Markdown(f"## {title}"))


header("1⃣ Default parsers — Japanese / Chinese (Simplified & Traditional) / Thai")


samples = {
   "Japanese (ja)":           ("今日は天気です。BudouXは機械学習を用いた改行整形ツールです。",
                               budoux.load_default_japanese_parser()),
   "Simplified Chinese":      ("今天是晴天。BudouX 是一个使用机器学习的换行整理工具。",
                               budoux.load_default_simplified_chinese_parser()),
   "Traditional Chinese":     ("今天是晴天。BudouX 是一個使用機器學習的換行整理工具。",
                               budoux.load_default_traditional_chinese_parser()),
   "Thai (th)":               ("วันนี้อากาศดีมากและฉันอยากออกไปเดินเล่นที่สวนสาธารณะ",
                               budoux.load_default_thai_parser()),
}
for title, (textual content, parser) in samples.objects():
   chunks = parser.parse(textual content)
   print(f"n• {title}")
   print(f"  uncooked   : {textual content}")
   print(f"  parsed:  '.be part of(chunks)    ({len(chunks)} phrases)")

header("2⃣ HTML translation with `translate_html_string`")


ja_parser = budoux.load_default_japanese_parser()
html_in = "今日は<b>とても天気</b>です。"
html_out = ja_parser.translate_html_string(html_in)
seen = html_out.change("u200b", "·")
print("Input  HTML :", html_in)
print("Output HTML :", html_out)
print("Visualised  :", seen)


demo_text = ("BudouXは機械学習を用いて、CJK言語の文章を意味のある"
            "フレーズに分割し、自然な位置で改行できるようにします。")
demo_html = ja_parser.translate_html_string(demo_text)
show(HTML(f"""
<div type="show:flex; hole:16px; font-family:'Hiragino Sans',sans-serif;">
 <div type="width:140px; border:2px stable #c33; padding:8px;">
    <b type="shade:#c33;"> Plain</b><br>{demo_text}
 </div>
 <div type="width:140px; border:2px stable #2a8; padding:8px;">
    <b type="shade:#2a8;"> BudouX</b><br>{demo_html}
 </div>
</div>
"""))


header("3⃣ Model introspection — options & weights")


model_dir = Path(budoux.__file__).dad or mum / "fashions"
print("Bundled fashions:", [p.name for p in model_dir.glob("*.json")])


with open(model_dir / "ja.json", encoding="utf-8") as f:
   ja_model = json.load(f)


print(f"nFeature classes in ja.json: {checklist(ja_model.keys())}")
complete = sum(len(v) for v in ja_model.values())
print(f"Total discovered options: {complete:,}")
for cat, feats in ja_model.objects():
   print(f"  • {cat:5s}  → {len(feats):,} options")


flat = [(cat, feat, w) for cat, d in ja_model.items() for feat, w in d.items()]
flat.type(key=lambda x: x[2], reverse=True)
print("nTop 5 options that vote 'BREAK HERE':")
for cat, feat, w in flat[:5]:
   print(f"  [{cat}] {feat!r}  → weight={w}")
print("nTop 5 options that vote 'DO NOT BREAK':")
for cat, feat, w in flat[-5:]:
   print(f"  [{cat}] {feat!r}  → weight={w}")

header("4⃣ Loading a customized mannequin with `budoux.Parser(mannequin)`")


neutered = {cat: {ok: 0 for ok in d} for cat, d in ja_model.objects()}
flat_parser = budoux.Parser(neutered)
print("All-zero mannequin output :", flat_parser.parse("今日は天気です。"))
print("Default mannequin output  :", ja_parser.parse("今日は天気です。"))


header("5⃣ Practical: customized separators, line-wrapping, JSON export")


def wrap_with_budoux(textual content, parser, max_width=12, sep="n"):
   traces, present = [], ""
   for phrase in parser.parse(textual content):
       if len(present) + len(phrase) > max_width and present:
           traces.append(present); present = phrase
       else:
           present += phrase
   if present: traces.append(present)
   return sep.be part of(traces)


novel = ("吾輩は猫である。名前はまだ無い。どこで生れたかとんと見当がつかぬ。"
        "何でも薄暗いじめじめした所でニャーニャー泣いていた事だけは記憶している。")
print("Wrapped at width 12:")
print(wrap_with_budoux(novel, ja_parser, max_width=12))


seg = {"textual content": novel, "phrases": ja_parser.parse(novel)}
print("nJSON payload (first 120 chars):", json.dumps(seg, ensure_ascii=False)[:120], "...")

header("6⃣ Performance benchmark")


big_text = novel * 200
t0 = time.perf_counter()
phrases = ja_parser.parse(big_text)
elapsed = time.perf_counter() - t0
print(f"Parsed {len(big_text):,} chars → {len(phrases):,} phrases "
     f"in {elapsed*1000:.1f} ms  ({len(big_text)/elapsed/1000:.0f}ok chars/sec)")


header("7⃣ Mini end-to-end coach (toy demo)")


training_lines = [
   "私は▁遅刻魔で、▁待ち合わせに▁いつも▁遅刻して▁しまいます。",
   "メールで▁待ち合わせ▁相手に▁一言、▁「ごめんね」と▁謝れば▁どうにか▁なると▁思って▁いました。",
   "海外では▁ケータイを▁持って▁いない。",
   "今日は▁とても▁いい▁天気です。",
   "明日は▁雨が▁降る▁かも▁しれません。",
   "週末は▁友達と▁映画を▁見に▁行きます。",
] * 20


SEP = "u2581"


def extract_features(s, i):
   def g(idx): return s[idx] if 0 <= idx < len(s) else ""
   feats = []
   for off in (-3,-2,-1,0,1,2):
       feats.append(f"U{off}:{g(i+off)}")
   for off in (-2,-1,0,1):
       feats.append(f"B{off}:{g(i+off)}{g(i+off+1)}")
   for off in (-1,0):
       feats.append(f"T{off}:{g(i+off)}{g(i+off+1)}{g(i+off+2)}")
   return feats


def make_examples(traces):
   X, y = [], []
   for line in traces:
       clear = line.change(SEP, "")
       breaks = set()
       j = 0
       for ch in line:
           if ch == SEP: breaks.add(j)
           else: j += 1
       for i in vary(1, len(clear)):
           X.append(extract_features(clear, i))
           y.append(1 if i in breaks else -1)
   return X, y


X, y = make_examples(training_lines)
print(f"Training examples: {len(X)}  (positives: {sum(1 for v in y if v==1)})")

def adaboost(X, y, rounds=80):
   n = len(y)
   w = [1/n]*n
   feat_set = sorted({f for fx in X for f in fx})
   fmap = [set(fx) for fx in X]
   model_rounds = []
   for r in vary(rounds):
       best_feat, best_err, best_pol = None, 1.0, 1
       for f in feat_set:
           err_pos = sum(w[i] for i in vary(n) if (f in fmap[i]) != (y[i]==1))
           err_neg = 1 - err_pos
           if err_pos < best_err: best_feat, best_err, best_pol = f, err_pos, +1
           if err_neg < best_err: best_feat, best_err, best_pol = f, err_neg, -1
       if best_err >= 0.5 - 1e-9: break
       eps = max(best_err, 1e-6)
       alpha = 0.5 * ( (1-eps)/eps ) ** 0.5
       new_w = []
       for i in vary(n):
           pred = best_pol if best_feat in fmap[i] else -best_pol
           new_w.append(w[i] * (0.5 if pred == y[i] else 2.0))
       s = sum(new_w); w = [x/s for x in new_w]
       model_rounds.append((best_feat, best_pol, alpha))
   return model_rounds


print("Training (this can be a toy coach — be affected person ~10s)...")
t0 = time.perf_counter()
rounds = adaboost(X, y, rounds=60)
print(f"Done in {time.perf_counter()-t0:.1f}s, {len(rounds)} stumps stored.")


appropriate = 0
for fx, label in zip(X, y):
   rating = sum(a if (f in fx) == (p==1) else -a for f,p,a in rounds)
   pred = 1 if rating > 0 else -1
   appropriate += (pred == label)
print(f"Training accuracy of toy mannequin: {appropriate/len(X)*100:.1f}%")
print(" For a manufacturing mannequin, use `scripts/prepare.py` from the BudouX repo with the matching characteristic extractor — this part is illustrative.")


header("8⃣ Real-world demo — slim column comparability")


paragraph = ("BudouXはGoogleが開発したオープンソースの改行ライブラリです。"
            "機械学習モデルを使って、文章を意味のあるフレーズに分割し、"
            "読みやすい位置でのみ改行が起こるようにします。"
            "依存関係がなく軽量なため、ウェブサイトやモバイルアプリに"
            "簡単に組み込むことができます。")
show(HTML(f"""
<div type="show:flex; hole:24px; font-family:'Hiragino Sans','Yu Gothic',sans-serif; font-size:15px;">
 <div type="flex:1; border:2px stable #c33; padding:12px; max-width:180px;">
   <b type="shade:#c33;">Without BudouX</b>
   <p type="line-height:1.7;">{paragraph}</p>
 </div>
 <div type="flex:1; border:2px stable #2a8; padding:12px; max-width:180px;">
   <b type="shade:#2a8;">With BudouX</b>
   <p type="line-height:1.7;">{ja_parser.translate_html_string(paragraph)}</p>
 </div>
</div>
<p type="font-size:12px;shade:#666;">Resize the browser/Colab pane to see the distinction extra clearly — BudouX by no means breaks a phrase mid-word.</p>
"""))


print("n Tutorial full. Try plugging BudouX output into your personal UI.")