initial revision

.gitignore

@@ -0,0 +1 @@
+source

README.md

@@ -1,3 +1,19 @@
 ---
-license: cc-by-4.0
+language: 
+- en
+pretty_name: "Cyberpunk"
+tags:
+- book-data
+license: cc-by-nc-4.0
 ---
+# Dataset - cyberpunk
+
+- **Developed by:** maldv
+- **License:** cc-by-nc-4.0
+- **Methodology:** Formatting book data by paragaph for training
+
+## Description
+
+Processing EBook data is much easier than having to deal with formatting long form book text.
+
+This is data artifacts from the processing a series of influential early cyberpunk books that I was able to find in epub format.  Enclosed is a jupyter notebook demonstrating the methodology.

cyberpunk.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:104047ddf465327d1b21427d53e0ea7990abe649c9bf7d9c95e1feaf2cbd8b80
+size 7220007

epub-processing.ipynb

@@ -0,0 +1,228 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from bs4 import BeautifulSoup, NavigableString, Tag\n",
+    "import ebooklib\n",
+    "from ebooklib import epub\n",
+    "import os\n",
+    "import re\n",
+    "from typing import Generator, List\n",
+    "\n",
+    "def parse_ebook_html(ebook_path: str, try_chapter : bool = False) -> Generator[tuple, None, None]:\n",
+    "    \"\"\"\n",
+    "    Parses the HTML content of an EPUB file, yielding only text content from each <p> block,\n",
+    "    while skipping specific elements with class 'calibre3' but considering valid text that follows.\n",
+    "\n",
+    "    Parameters:\n",
+    "    - ebook_path (str): The path to the EPUB file.\n",
+    "    - try_chapter (bool): If True, the first paragraph of each chapter will be used to determine the chapter title.\n",
+    "\n",
+    "    Returns:\n",
+    "    - text_generator (Generator[tuple, None, None]): A generator yielding text content.\n",
+    "    \"\"\"\n",
+    "    book = epub.read_epub(ebook_path)\n",
+    "    basename = os.path.basename(ebook_path)\n",
+    "    noext = os.path.splitext(basename)[0]\n",
+    "    chapter_idx = 0\n",
+    "    paragraph_idx = 0\n",
+    "    cumsum_word_count = 0\n",
+    "    for item in book.get_items_of_type(ebooklib.ITEM_DOCUMENT):\n",
+    "        content = item.get_content().decode('utf-8')\n",
+    "        results = list(html_tokenizer(content, try_chapter))\n",
+    "        if len(results) == 0:\n",
+    "            continue\n",
+    "        chapter_idx += 1\n",
+    "        for row in results:\n",
+    "            if len(row[1]) == 0:\n",
+    "                continue\n",
+    "            paragraph_idx += 1\n",
+    "            word_count = len((row[1]))\n",
+    "            cumsum_word_count += word_count\n",
+    "            row = [noext, paragraph_idx, chapter_idx] + list(row[:]) + [word_count, cumsum_word_count]\n",
+    "            yield tuple(row)\n",
+    "\n",
+    "def html_tokenizer(html_content: str, try_chapter) -> Generator[tuple, None, None]:\n",
+    "    \"\"\"\n",
+    "    Generator function to tokenize HTML content, yielding text content from each <p> block.\n",
+    "\n",
+    "    Parameters:\n",
+    "    - html_content (str): The HTML content to be tokenized.\n",
+    "    - try_chapter (bool): If True, the first paragraph of each chapter will be used to determine the chapter title.\n",
+    "\n",
+    "    Yields:\n",
+    "    - text_generator (Generator[tuple, None, None]): A generator yielding text content. \n",
+    "    \"\"\"\n",
+    "    soup = BeautifulSoup(html_content, 'html.parser')\n",
+    "    fix_quote = re.compile(r'“|”|»|«')\n",
+    "    fix_threedot = re.compile(r'…')\n",
+    "    fix_bars = re.compile(r'\\|\\s*\\|')\n",
+    "\n",
+    "    def extract_and_yield_text(element, accumulated_texts: List[str]):\n",
+    "        if isinstance(element, NavigableString):\n",
+    "            accumulated_texts.append(str(element))\n",
+    "        elif isinstance(element, Tag):\n",
+    "            if element.name == 'a' and 'calibre3' in element.get('class', []):\n",
+    "                # Skip processing the <a class=\"calibre3\"> tag itself, but not its siblings\n",
+    "                #print('skipping', element)\n",
+    "                return\n",
+    "            if element.name == 'span' and 'italic' in element.get('class', []):\n",
+    "                # Append italic text directly to the accumulated_texts list without yielding\n",
+    "                accumulated_texts.append(element.get_text())\n",
+    "            else:\n",
+    "                # Recursively process all children, including those following skipped elements\n",
+    "                for child in element.children:\n",
+    "                    extract_and_yield_text(child, accumulated_texts)\n",
+    "\n",
+    "    chapter = None\n",
+    "    for i, p_tag in enumerate(soup.find_all('p')):\n",
+    "        accumulated_texts = []\n",
+    "        # if p's class is calibre14, skip it because it's metadata\n",
+    "        if 'calibre14' in p_tag.get('class', []):\n",
+    "            #print('skipping', i)\n",
+    "            #continue\n",
+    "            pass\n",
+    "        else:\n",
+    "            #print('processing', i)\n",
+    "            if i == 0 and try_chapter:\n",
+    "                # Instead of processing, this contains our chapter and title\n",
+    "                markers = []\n",
+    "                for span in p_tag.find_all('span', class_='bold'):\n",
+    "                    markers.append(span.get_text())\n",
+    "\n",
+    "                if len(markers) >= 2:\n",
+    "                    chapter = ' '.join(markers)\n",
+    "                continue\n",
+    "                    \n",
+    "        extract_and_yield_text(p_tag, accumulated_texts)\n",
+    "        # if our text is '| |', skip it\n",
+    "        if '| |' in ' '.join(accumulated_texts):\n",
+    "            continue\n",
+    "        text = ' '.join([text.strip() for text in accumulated_texts if text.strip()])\n",
+    "        text = text.replace(u'\\xa0', u' ')\n",
+    "        text = fix_quote.sub(u'\"', text)\n",
+    "        text = fix_threedot.sub(u'...', text)\n",
+    "        text = fix_bars.sub(u'', text)\n",
+    "        text = text.strip()\n",
+    "        # If the first character is a capital letter, then a space, followed by more capital letters, it is likely the beginning of a chapter and needs to have the space removed\n",
+    "        if len(text) == 0:\n",
+    "            continue\n",
+    "        elif len(text) > 2 and text[0].isupper() and text[1] == ' ' and text[2].isupper():\n",
+    "            text = text[0] + text[2:]\n",
+    "        yield chapter, text\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Processing cryptonomicon\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/mnt/biggy/ai/notebook/jupyterenv/lib/python3.10/site-packages/ebooklib/epub.py:1395: UserWarning: In the future version we will turn default option ignore_ncx to True.\n",
+      "  warnings.warn('In the future version we will turn default option ignore_ncx to True.')\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Clipping (1, 2)\n",
+      "Processing neuromancer\n",
+      "Clipping (1, 49)\n",
+      "Processing burningchrome\n",
+      "Processing snowcrash\n",
+      "Clipping (1, 80)\n",
+      "Clipping (4590, 4636)\n",
+      "Processing quicksilver\n",
+      "Processing monalisa\n",
+      "Processing theconfusion\n",
+      "Processing systemofworld\n",
+      "Clipping (261, 331)\n",
+      "Saved 41535 paragraphs to cyberpunk.parquet\n"
+     ]
+    }
+   ],
+   "source": [
+    "from glob import glob\n",
+    "import pandas as pd\n",
+    "\n",
+    "# special rules.\n",
+    "# cryptonomicon requires try_chapter=True, and needs '| |' removed\n",
+    "special = {\n",
+    "    'cryptonomicon': {'try_chapter': True, 'clip': [(1,2),], 'drop': [106]},\n",
+    "    'neuromancer' : {'clip': [(1,49)]},\n",
+    "    'burningchrome' : { \"try_chapter\": True, 'drop': [1]},\n",
+    "    'snowcrash'     : { \"try_chapter\": True, 'clip': [(1, 80), (4590, 4636)]},\n",
+    "    'quicksilver' : {},\n",
+    "    'monalisa' : {},\n",
+    "    'theconfusion' : {},\n",
+    "    'systemofworld' : {'clip': [(261, 331)]},\n",
+    "}\n",
+    "\n",
+    "all_books = pd.DataFrame([], columns=['book_name', 'paragraph_ix', 'chapter_ix', 'chapter_title', 'text', 'word_count', 'cumsum_word_count'])\n",
+    "for book in glob('source/*.epub'):\n",
+    "    name = os.path.splitext(os.path.basename(book))[0]\n",
+    "    print(f\"Processing {name}\")\n",
+    "    try_chapter = False\n",
+    "    clips = []\n",
+    "    drops = []\n",
+    "    if name in special:\n",
+    "        config = special[name]\n",
+    "        try_chapter = config.get('try_chapter', False)\n",
+    "        if 'clip' in config:\n",
+    "            clips = config['clip']\n",
+    "        if 'drop' in config:\n",
+    "            drops = config['drop']\n",
+    "\n",
+    "    lines = parse_ebook_html(book, try_chapter=try_chapter)\n",
+    "    new_frame = pd.DataFrame(lines, columns=['book_name', 'paragraph_ix', 'chapter_ix', 'chapter_title', 'text', 'word_count', 'cumsum_word_count'])\n",
+    "    for drop in drops:\n",
+    "        new_frame = new_frame[new_frame['chapter_ix'] != drop]\n",
+    "    for clip in clips:\n",
+    "        # we want to remove the paragraph id's that are in the clip range, inclusive\n",
+    "        print(f\"Clipping {clip}\")\n",
+    "        idxr = range(clip[0], clip[1] + 1)\n",
+    "        new_frame = new_frame[~new_frame['paragraph_ix'].isin(idxr)]\n",
+    "            \n",
+    "    all_books = pd.concat([all_books, new_frame.copy()], ignore_index=True)\n",
+    "all_books.to_parquet('cyberpunk.parquet')\n",
+    "print(f\"Saved {len(all_books)} paragraphs to cyberpunk.parquet\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "jupyterenv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

requirements.txt

@@ -0,0 +1,4 @@
+beautifulsoup4==4.12.2
+EbookLib==0.18
+pandas==2.1.3
+pyarrow==15.0.0