Initial commit

core/epub_reader.py

@@ -0,0 +1,25 @@
+from ebooklib import epub
+from ebooklib import ITEM_DOCUMENT
+from bs4 import BeautifulSoup
+
+CHAPTER_TAGS = {"h1", "h2", "h3", "h4"}
+
+def read_epub(path):
+    book = epub.read_epub(path)
+    chapters = []
+    for item in book.get_items():
+        if item.get_type() == ITEM_DOCUMENT:
+            soup = BeautifulSoup(item.get_content(), "html.parser")
+            chapters.append(soup.get_text())
+    return chapters
+
+def read_epub_flat(path):
+    return "\n\n".join(read_epub(path))
+
+def read_epub_metadata(path):
+    book = epub.read_epub(path)
+    title = book.get_metadata('DC', 'title')
+    author = book.get_metadata('DC', 'creator')
+    title_str = title[0][0] if title else None
+    author_str = author[0][0] if author else None
+    return title_str, author_str

core/scene_splitter.py

@@ -0,0 +1,62 @@
+import re
+
+JUNK_PATTERNS = [
+    r'all rights reserved',
+    r'copyright\s*©',
+    r'isbn[\s\-]',
+    r'published by',
+    r'first published',
+    r'printed in',
+    r'no part of this',
+    r'table of contents',
+    r'this is a work of fiction',
+    r'any resemblance to',
+]
+JUNK_RE = re.compile('|'.join(JUNK_PATTERNS), re.IGNORECASE)
+
+def _is_junk(text):
+    if len(text.split()) < 30:
+        return True
+    if JUNK_RE.search(text):
+        return True
+    return False
+
+def split_scenes(chapters, min_sentences=3, max_sentences=6):
+    scenes = []
+    for chapter_text in chapters:
+        paragraphs = [p.strip() for p in re.split(r'\n{2,}', chapter_text) if p.strip()]
+        buf = []
+        buf_sentences = 0
+
+        for para in paragraphs:
+            sentences = re.split(r'(?<=[.!?])\s+', para.strip())
+            sentences = [s for s in sentences if s]
+
+            for sentence in sentences:
+                buf.append(sentence)
+                buf_sentences += 1
+
+                if buf_sentences >= max_sentences:
+                    candidate = " ".join(buf)
+                    if not _is_junk(candidate):
+                        scenes.append(candidate)
+                    buf = []
+                    buf_sentences = 0
+
+            if buf_sentences >= min_sentences:
+                candidate = " ".join(buf)
+                if not _is_junk(candidate):
+                    scenes.append(candidate)
+                buf = []
+                buf_sentences = 0
+
+        if buf:
+            candidate = " ".join(buf)
+            if scenes and _is_junk(candidate):
+                pass
+            elif scenes:
+                scenes[-1] = scenes[-1] + " " + candidate
+            elif not _is_junk(candidate):
+                scenes.append(candidate)
+
+    return [s for s in scenes if len(s.strip()) > 20]