論語The Analects of Confucius

Validation marks are turned off. You can turn them back on in .

Sources and validation

How this was made

Some of what this site shows is reproduced from published books. Much of the rest was made by machine from the Chinese commentaries. Every piece of the machine-made data carries a small mark that says whether anyone has checked it against its source, and a tap on the mark shows who checked it, when, and what they found.

Published sources

The Chinese text of each passage and chapter, and the translations printed under it, come from published editions. They're named rather than validated: the page mark opens the edition.

  • The received text of the Analects (論語), and the 王弼 text of the Dao De Jing (王弼本).
  • James Legge, The Chinese Classics, vol. 1 (1861), and The Tâo Teh King, in The Sacred Books of the East, vol. 39 (1891).
  • Paul Carus, The Canon of Reason and Virtue (1913).
  • The translations quoted with an episode, such as Peimin Ni's and Edward Slingerland's, each credited where it's shown.

Elliott Bernstein's own episode guides and translations are his work. They carry a green mark in his name, and machine checks for possible inaccuracies can be added to them later.

What was made by machine

The English summaries and translations of the commentaries were made by machine from the Chinese: for the Analects mostly from 程樹德's 論語集釋 (1943), and for the Dao De Jing from 河上公, 王弼, 焦竑's 老子翼, 憨山德清, 王夫之 and 陸德明. That covers the disputes and their readings, the verdicts, the notes, the lines that say who a person was or what a work is, the one-line hooks, and the senses on the word pages.

The camps, the divergences between commentators and every count of them (how often two people read alike or differently, who is an outlier) are this site's own inferences from that data. No commentator grouped himself into a camp, so treat them as a way into the commentary rather than a claim about it.

What the marks mean

Not validated yet
Nobody has checked it against its source.
Checked by a machine
A model checked it against the Chinese source and found it accurate.
Checked by a person, with a doubt
A named person checked it and kept it, but noted a doubt. The records say what the doubt is and which source it follows.
Checked by a person
A named person checked it and confirmed it.
A concern was raised
A check found a problem that hasn't been resolved yet. The records say what it is.

The least confident record wins: an open concern makes a mark red, whatever else has been said about it. A check counts only for the exact text it checked. When the text changes, the old records stay in its history marked "changed since validated", and the mark goes back to gray until somebody checks the new version. A concern is closed only by a later record that says so, such as a second opinion that finds the item sound.

Who checks

The machine checks so far are by Claude Opus 5.5, a model made by Anthropic. Each check is independent of the run that made the item: the checker has the Chinese source in front of it and doesn't see any earlier verdicts. Each record names the model and version, the date, the verdict, the part of the source it was checked against, and a short comment.

The only person who has validated anything so far is Elliott Bernstein, for his own episode guides. Other people may validate items later, and their records will carry their names.

Where the checks stand

Every machine-made item of the Analects has been checked against its source. Each check was made without seeing any earlier verdict, and every concern got a second opinion from a separate check. The items with concerns that held were revised and checked again, three times over.

The AnalectsItemsMachinePersonPerson, doubt notedConcern
Dispute questions1,0871,087000
Camp groupings1,0871,087000
Readings2,7932,793000
Quotations10,35010,350000
Verdicts330330000
Notes3,1873,187000
People1,2761,276000
Works975975000
Hooks6565000
Word links555555000
Why a character has no link511510100
Word definitions in a passage21,79621,7820140
Word senses across the book3,5443,544000
Episode guides2802800
Grammar pages1601600

Across the site the records cover 65,099 items: 64,515 checked by a machine, 46 checked by a person, 14 checked by a person with a doubt noted, 0 with an open concern and 524 changed since they were checked. Most of the Dao De Jing hasn't been checked yet, so its marks are still gray.

Reporting a problem

Every opened dispute and note has a "Report a problem" link, and the apps have a report button. Reports go to a review queue. They don't change a mark by themselves: a report is reviewed, and the mark changes only when the review adds a record to the item's history.