Upload scrape_kawmuthu.py with huggingface_hub

scrape_kawmuthu.py

@@ -0,0 +1,192 @@
+"""
+Scraper for kawmuthu.blogspot.com — Sinhala poems by Kalani Ruwanpathirana
+Run on your local machine or in Kaggle (internet must be ON)
+
+Usage:
+    pip install requests beautifulsoup4
+    python scrape_kawmuthu.py
+
+Output:
+    kawmuthu_poems_raw.jsonl       — raw scraped poems
+    kawmuthu_poems_training.jsonl  — Alpaca format ready for training
+"""
+
+import requests
+from bs4 import BeautifulSoup
+import json
+import re
+import time
+
+# ── Helpers ─────────────────────────────────────────────────
+
+def is_sinhala(text):
+    sinhala_chars = sum(1 for c in text if '\u0D80' <= c <= '\u0DFF')
+    return sinhala_chars > 30
+
+def clean_poem(text):
+    text = re.sub(r'@[A-Za-z\s\.]+', '', text)       # remove @Author
+    text = re.sub(r'©[A-Za-z\s\.]+', '', text)       # remove ©Author
+    text = re.sub(r'-\d{4}\.\d{2}\.\d{2}', '', text) # remove date
+    text = re.sub(r'\n{3,}', '\n\n', text)            # reduce blank lines
+    text = re.sub(r'[ \t]+\n', '\n', text)            # trailing spaces
+    return text.strip()
+
+def get_all_post_urls(base_url):
+    """Crawl all pages and collect post URLs."""
+    urls = set()
+    page_url = base_url
+    page_num = 1
+
+    while page_url:
+        print(f'  Page {page_num}: {page_url}')
+        try:
+            resp = requests.get(page_url, timeout=15)
+            soup = BeautifulSoup(resp.text, 'html.parser')
+
+            for a in soup.find_all('a', href=True):
+                href = a['href']
+                if 'kawmuthu.blogspot.com/20' in href:
+                    # strip query params
+                    href = href.split('?')[0]
+                    urls.add(href)
+
+            older = soup.find('a', id='Blog1_blog-pager-older-link')
+            page_url = older['href'] if older else None
+            page_num += 1
+            time.sleep(1)
+
+        except Exception as e:
+            print(f'  Error on page {page_num}: {e}')
+            break
+
+    return list(urls)
+
+def scrape_post(url):
+    """Scrape a single blog post and return poem dict or None."""
+    try:
+        resp = requests.get(url, timeout=15)
+        soup = BeautifulSoup(resp.text, 'html.parser')
+
+        # Title
+        title_tag = (soup.find('h3', class_='post-title') or
+                     soup.find('h1', class_='post-title') or
+                     soup.find('h3', class_='entry-title'))
+        title = title_tag.get_text(strip=True) if title_tag else ''
+
+        # Labels
+        labels = [a.get_text(strip=True) for a in soup.find_all('a', rel='tag')]
+
+        # Body
+        body = (soup.find('div', class_='post-body') or
+                soup.find('div', class_='entry-content'))
+        if not body:
+            return None
+
+        for tag in body.find_all(['img', 'script', 'style']):
+            tag.decompose()
+
+        text = body.get_text('\n')
+        text = clean_poem(text)
+
+        if not is_sinhala(text):
+            return None
+        if len(text) < 60:
+            return None
+
+        return {
+            'title': title,
+            'content': text,
+            'labels': labels,
+            'url': url,
+            'author': 'Kalani Ruwanpathirana',
+            'source': 'kawmuthu.blogspot.com',
+            'type': 'poem',
+        }
+    except Exception as e:
+        print(f'  Error: {url} — {e}')
+        return None
+
+# ── Training format ──────────────────────────────────────────
+
+def make_training_examples(poem):
+    """Generate 5 Alpaca training examples per poem."""
+    content = poem['content']
+    title   = poem['title']
+    author  = poem['author']
+    labels  = poem['labels']
+
+    # Detect theme from labels
+    theme = labels[0] if labels else 'සිංහල'
+    theme_si = {
+        'love': 'ආදර', 'life': 'ජීවිත', 'nature': 'ස්වභාව',
+        'baby': 'ළදරු', 'night': 'රාත්‍රී', 'moon': 'සඳ',
+        'beauty': 'සෞන්දර්ය', 'sadness': 'දුක',
+    }.get(theme.lower(), 'සිංහල')
+
+    TEMPLATE = (
+        'Below is an instruction that describes a task. '
+        'Write a response that appropriately completes the request.\n\n'
+        '### Instruction:\n{instruction}\n\n'
+        '### Response:\n{response}'
+    )
+
+    instructions = [
+        f'සිංහල කවියක් ලියන්න',
+        f'Write a Sinhala poem',
+        f'{theme_si} ගැන සිංහල කවියක් ලියන්න',
+        f"'{title}' වැනි සිංහල කවියක් ලියන්න",
+        f'{author} ගේ ශෛලියෙන් සිංහල කවියක් ලියන්න',
+    ]
+
+    examples = []
+    for inst in instructions:
+        examples.append({
+            'text': TEMPLATE.format(instruction=inst, response=content),
+            'instruction': inst,
+            'output': content,
+            'source': poem['source'],
+            'type': 'poem',
+            'title': title,
+        })
+    return examples
+
+# ── Main ─────────────────────────────────────────────────────
+
+if __name__ == '__main__':
+    BASE = 'https://kawmuthu.blogspot.com/'
+
+    print('Step 1: Collecting post URLs...')
+    urls = get_all_post_urls(BASE)
+    print(f'Found {len(urls)} post URLs\n')
+
+    print('Step 2: Scraping poems...')
+    poems = []
+    for i, url in enumerate(urls):
+        poem = scrape_post(url)
+        if poem:
+            poems.append(poem)
+            print(f'  [{i+1}/{len(urls)}] ✓ {poem["title"][:50]}')
+        else:
+            print(f'  [{i+1}/{len(urls)}] - skipped')
+        time.sleep(0.5)
+
+    print(f'\nScraped {len(poems)} valid poems')
+
+    # Save raw
+    with open('kawmuthu_poems_raw.jsonl', 'w', encoding='utf-8') as f:
+        for p in poems:
+            f.write(json.dumps(p, ensure_ascii=False) + '\n')
+    print('Saved kawmuthu_poems_raw.jsonl')
+
+    # Save training format
+    all_examples = []
+    for poem in poems:
+        all_examples.extend(make_training_examples(poem))
+
+    with open('kawmuthu_poems_training.jsonl', 'w', encoding='utf-8') as f:
+        for ex in all_examples:
+            f.write(json.dumps(ex, ensure_ascii=False) + '\n')
+
+    print(f'Saved kawmuthu_poems_training.jsonl ({len(all_examples)} training examples)')
+    print(f'\nDone! Combine with your existing data:')
+    print('  cat combined_training.jsonl kawmuthu_poems_training.jsonl > final_training.jsonl')