Add dl archive.org script for downloading indices

README.md

@@ -2,7 +2,6 @@
 license: mit
 task_categories:
   - text-generation
-  - text2text-generation
 language:
   - en
 tags:
@@ -19,18 +18,18 @@ size_categories:
 
 # A dataset of pre-1950 English text
 
-For training LLMs on old science, in order to quiz them about current
-capabilities and developments
+This dataset is for training LLMs on old science, in order to quiz them about
+current capabilities and developments. The goal is to explore how much
+prompting an old-LLM would require in order to "invent" modern technology, with the
+hope that this will inform how to get current LLMs to truly invent
+next-generation technology.
 
-Possible data sources:
+# Data sources to investigate
 
-- www.chroniclingamerica.loc.gov: https://chroniclingamerica.loc.gov/lccn/sn86072192/1897-08-01/ed-1/seq-1/
-- www.gutenberg.org/
-- https://www.newspapers.com (https://www.newspapers.com/paper/the-45th-division-news/29237/) (licensing unclear, requires sign-in)
 - wikipedia looks like it's got a big list of newspaper archives: https://en.wikipedia.org/wiki/Wikipedia:List_of_online_newspaper_archives
 - also see https://github.com/haykgrigo3/TimeCapsuleLLM
 
-# Data Sources
+# Data Sources in use
 
 ## Project Gutenberg
 
@@ -70,18 +69,24 @@ Information: https://chroniclingamerica.loc.gov/ocr/
 
 JSON listing of files: https://chroniclingamerica.loc.gov/ocr.json
 
-Download all:
+Download the full dataset, one archive at a time (total size is 2 115 GB):
 
 ```
 uv run src/download_chronicling_america.py
 ```
 
-Find all directories indicating content after 1950 and delete:
+Conveniently, they're all organised by date, so we can find all directories
+indicating content after 1950 and delete them:
 
 ```
 find -E data/chronicling-america -type d -regex '.*/(1950|19[5-9][0-9]|20[0-9]{2})$' -exec rm -rf {} +
 ```
 
+TODO the resulting files are pretty bad. The OCR has many many artefacts, and
+not all of them are obvious how to fix, since the source scans/images aren't
+available apparently. Not sure how to fix these without using modern LLMs and
+potentially infecting the dataset.
+
 ## Biodiversity Heritage Library
 
 60+ million pages of OCR content (~41 GB)
@@ -122,6 +127,23 @@ exactly how to do so.
 
 https://data.uspto.gov/apis/getting-started
 
+## Hathi Trust
+
+> HathiTrust was founded in 2008 as a not-for-profit collaborative of academic
+> and research libraries now preserving 18+ million digitized items in the
+> HathiTrust Digital Library. We offer reading access to the fullest extent
+> allowable by U.S. and international copyright law, text and data mining tools
+> for the entire corpus, and other emerging services based on the combined
+> collection.
+
+Looks like it has a lot of information, although this might all just be
+duplicated from the data available in the Internet Archive. Also it's less
+easy to download, Google Books has some pretty restrictive licensing
+
+https://babel.hathitrust.org/cgi/pt?id=mdp.39015082239875&seq=26&format=plaintext
+
+[Advanced search URL](https://babel.hathitrust.org/cgi/ls?lmt=ft&a=srchls&adv=1&c=148631352&q1=*&field1=ocr&anyall1=all&op1=AND&yop=before&pdate_end=1949&facet_lang=language008_full%3AEnglish&facet_lang=language008_full%3AEnglish%2C+Middle+%281100-1500%29&facet_lang=language008_full%3AEnglish%2C+Old+%28ca.+450-1100%29&facet_format=format%3ADictionaries&facet_format=format%3AEncyclopedias&facet_format=format%3AJournal&facet_format=format%3AManuscript&facet_format=format%3ANewspaper&facet_format=format%3ABiography&facet_format=format%3ABook)
+
 # Cleaning up
 
 List all file containing at least 1% lines of non-English characters

src/download_archive_dot_org.py

@@ -0,0 +1,91 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = ["tqdm"]
+# ///
+import csv
+import urllib.parse
+import urllib.request
+from pathlib import Path
+from datetime import date, timedelta
+from tqdm import tqdm
+
+BASE_URL = "https://archive.org/advancedsearch.php"
+FIELDS = ['date', 'identifier', 'item_size']
+ROWS_PER_PAGE = 1000
+
+OUT_ROOT = Path("data/archive-dot-org/indices")
+BY_YEAR = OUT_ROOT / "by_year"
+BY_DAY = OUT_ROOT / "by_day"
+BY_YEAR.mkdir(parents=True, exist_ok=True)
+BY_DAY.mkdir(parents=True, exist_ok=True)
+
+def parse_csv(body: str) -> list[dict]:
+    lines = body.strip().splitlines()
+    if not lines or lines[0].startswith('<!DOCTYPE html>'):
+        return []
+    reader = csv.DictReader(lines)
+    return list(reader)
+
+def fetch_range(start: str, end: str) -> list[dict]:
+    all_rows = []
+    page = 1
+    while True:
+        q = f'mediatype:(texts) AND language:(English) AND date:[{start} TO {end}]'
+        params = {
+            'q': q,
+            'fl[]': FIELDS,
+            'sort[]': 'date asc',
+            'rows': str(ROWS_PER_PAGE),
+            'page': str(page),
+            'output': 'csv',
+        }
+        url = BASE_URL + '?' + urllib.parse.urlencode(params, doseq=True)
+        try:
+            with urllib.request.urlopen(url) as r:
+                body = r.read().decode('utf-8')
+        except Exception as e:
+            print(f"  Failed {start}–{end} page {page}: {e}")
+            break
+        rows = parse_csv(body)
+        if not rows:
+            break
+        all_rows.extend(rows)
+        if len(rows) < ROWS_PER_PAGE:
+            break
+        page += 1
+    return all_rows
+
+# 1. Pre-1900: yearly
+pbar = tqdm(range(1600, 1900), desc="Year bins")
+for year in pbar:
+    out_file = BY_YEAR / f"{year}.csv"
+    if out_file.exists():
+        continue
+    pbar.set_description(f"Fetching {year}")
+    rows = fetch_range(f"{year}-01-01", f"{year}-12-31")
+    if rows:
+        pbar.set_description(f"Fetching {year} (writing {len(rows)} rows)")
+        with out_file.open("w", newline='', encoding="utf-8") as f:
+            writer = csv.DictWriter(f, fieldnames=FIELDS)
+            writer.writeheader()
+            writer.writerows(rows)
+
+# 2. 1900–1949: daily
+start_date = date(1900, 1, 1)
+end_date = date(1950, 1, 1)
+cur = start_date
+pbar = tqdm(total=(end_date - start_date).days, desc="Daily bins")
+while cur < end_date:
+    day_str = cur.isoformat()
+    out_file = BY_DAY / f"{day_str}.csv"
+    if not out_file.exists():
+        pbar.set_description("Fetching {day_str}")
+        rows = fetch_range(day_str, day_str)
+        if rows:
+            pbar.set_description("Fetching {day_str} (writing {len(rows)} rows)")
+            with out_file.open("w", newline='', encoding="utf-8") as f:
+                writer = csv.DictWriter(f, fieldnames=FIELDS)
+                writer.writeheader()
+                writer.writerows(rows)
+    cur += timedelta(days=1)
+    pbar.update(1)

src/download_chronicling_america.py

@@ -37,10 +37,9 @@ def main() -> None:
         out_dir = DEST_ROOT / name.removesuffix(".tar.bz2")
 
         if out_dir.exists():
+            print(f"Directory exists, skipping {out_dir}")
             continue
-
-
-        if not out_dir.exists():
+        else:
             if not archive_path.exists():
                 print(f"downloading {name}")
                 urllib.request.urlretrieve(url, archive_path)
@@ -48,8 +47,6 @@ def main() -> None:
             with tarfile.open(archive_path, "r:bz2") as tar:
                 tar.extractall(out_dir)  # simple, unsafe but short
             os.remove(archive_path)
-        else:
-            print(f"Directory exists, skipping {out_dir}")
 
     print("done")