ctext Wiki Explorer

A single web page that lists every book in the Wiki of the Chinese Text Project (ctext.org) with an English title.

Open the site: https://orpheus-21.github.io/ctext-wiki-explorer/

What it does

The ctext Wiki holds about 49,500 digitized editions of pre-modern Chinese books. The ctext site shows them page by page and only in Chinese. This project puts all of them on one page in English. The page groups repeated editions of one title into one entry, so the page shows about 38,600 books.

Each entry shows:

• An English title. Most English titles are machine translations. A small MT mark shows these titles.
• The author, with the name in pinyin.
• The period, the subject from the old imperial library catalogues, and the kind of book.
• A link to each edition on ctext.org. The text on ctext.org is in Chinese.
• A link to the Data Wiki record and to English Wikipedia, when they exist.

On the page, you can:

• Search in English. The search looks at the title, the author, the subject, the period and the kind of book. Every word must match.
• Filter by subject, by period and by kind of book. Each filter shows the number of matching books.
• Show only famous books. A famous book or its author has an English Wikipedia article.
• Click an author to see all books by that author.
• Click "Surprise me" to get a random book from the current list.
• Sort by the amount of information, at random, from A to Z, or by the number of editions.

The page has a second tab, "Statistics". The tab shows:

• Counts for this site: books, editions, authors, the sources of the English titles, and the number of corrections in review_changes.tsv.
• Figures for the whole Chinese Text Project, with the date of the last ctext update.
• Bar charts of the books by period, by kind and by subject, the authors with the most books, and the books with the most editions. A click on a bar opens the book list with that filter.

The address index.html#stats opens the Statistics tab directly.

The page contains no Chinese characters.

Requirements

• Python 3. The scripts were tested with Python 3.14.
• The Python package pypinyin, version 0.55 or later. Only build.py needs it.
• An internet connection for the first build. build.py downloads about 12 MB from ctext.org.
• To make or check English titles: the claude command-line tool, logged in. Only translate.py and review.py need it.
• A web browser to open the page.

Install

1. Clone the repo:

   git clone https://github.com/Orpheus-21/ctext-wiki-explorer.git
   cd ctext-wiki-explorer
   

2. Make a virtual environment:

   python3 -m venv .venv
   

3. Install the dependency:

   .venv/bin/pip install -r requirements.txt
   

Usage

Open the page

Open the site address in a web browser. GitHub Pages serves index.html from the main branch.

To use the page offline, open index.html in a web browser. The repo contains a built copy.

Build the page

Run the build:

.venv/bin/python build.py

The script writes index.html. It also writes cache/books.json, which translate.py and review.py read.

To get new data from ctext.org, delete the cache/ folder. Then run the build again.

Translate titles

translate.py gives an English title to each title that has none. Run build.py once before you run translate.py.

Command	Result
python3 translate.py --test	Translates one sample batch. Shows the token use and 30 samples.
python3 translate.py 5	Translates the next 5 batches of 500 titles.
python3 translate.py	Translates all titles that have no English title.

The script appends each result to titles_en.tsv. If a run stops, run the same command again. The script continues with the remaining titles.

Check titles

review.py compares each English title with its Chinese title. It changes only the titles that are wrong.

python3 review.py 2

The number is the count of batches of 500 titles. The default is 1. The script writes the corrections into titles_en.tsv and adds each change to review_changes.tsv.

After a translation run or a check, run build.py again to put the new titles on the page.

Files

File	Content
build.py	Downloads the data, joins it, and writes index.html.
template.html	The page layout, styles and script. build.py puts the data into it.
translate.py	Makes English titles in batches.
review.py	Checks and corrects English titles in batches.
titles_en.tsv	One line per title: the Chinese title, a tab, the English title.
review_changes.tsv	One line per correction: the Chinese title, the old English title, the new English title.
index.html	The built page.
cache/	Downloaded data and run state. Git ignores this folder.

How it works

build.py uses three sources from ctext.org:

1. The API function gettexttitles. It gives the title and the identifier of each Wiki item.
2. The Data Wiki dump in RDF format. It gives works, authors, dynasties, dates and catalogue subjects.
3. The API function getstats. It gives the figures for the whole Chinese Text Project on the Statistics tab.

The script joins each Wiki item to a work in the dump. It uses the work identifier first and the exact title second. Items that have the same title become one entry. The script finds the kind of book from the last characters of the Chinese title. It finds the period from the dynasty or the dates of the author.

The English title of an entry comes from one of three sources, in this order:

1. The name of the English Wikipedia article of the work.
2. The line in titles_en.tsv.
3. A pinyin form of the Chinese title. The page marks these titles with ROM.

The author name comes from the English Wikipedia article of the author. If no article exists, the script makes the name from pinyin.

The script then removes all Chinese text from the data and puts the data into template.html. The build stops with an error if a Chinese character is in the result.

translate.py and review.py send each batch to the claude command-line tool as one new request. The tool must return the Chinese title with each English title. A script keeps a reply line only when the returned title matches a title in the batch. This rule stops a skipped line from putting a translation on the wrong title. Titles are compared after Unicode NFKC normalization and with whitespace removed.

Limits

• The English titles are machine translations. Some are wrong. review.py corrects a part of them.
• About 23 percent of the books have an author in the data. About 19 percent have a subject.
• Automatic pinyin can choose the wrong reading of a character with two readings.

License

The code is licensed under the GNU General Public License, version 3 or any later version. See the LICENSE file.

The data comes from the Chinese Text Project. The Data Wiki data is licensed under the Creative Commons Attribution, NonCommercial, ShareAlike 3.0 license (CC BY-NC-SA 3.0). This license also applies to index.html, titles_en.tsv and review_changes.tsv, because they contain this data. You must credit the Chinese Text Project, you must not use the data for commercial purposes, and you must share changes under the same license.

This project is not affiliated with the Chinese Text Project.