Chapter 14. Integrity of Materials, and a Final Greeting

Opening Part 5. Over thirteen installments we have seen how a work’s materials are made: the place and framework of the work in Part 1; the inside of the work folder in Part 2; the thirty-five kinds one by one in Part 3; character notes and supporting files in Part 4. The final installment holds two things together: verification, by which all these materials become one integral reference book, and the close of this fourteen-part series. They belong together because verification is the materials’ final finishing; only when it ends is the folder complete, and the completion of the folder is the completion of our work.

Materials Reflecting One Another

When thirty-five kinds of material are made, notes for thirty-four characters are written, a 156-line event index is organized, and eighteen chunk-by-chunk tables are complete, one question follows automatically: do all these materials truly form one consistent reference book?

Is the same character’s name written the same way everywhere? Has the same word been rendered differently in Korean? Do all prior-event references point to real events? No one can hold every material in their head, so materials must reflect materials. That is Stage 6: final cross-verification.

Seven Check Items

We inspect seven places.

First, agreement between character notes and chunk tables. Does every person appearing in a chunk table have a note? Has a person with a note ever been omitted from the table? If a person appears in one material but not another, coherence breaks.

Second, event-number existence on both sides. Every event number in the event master must be in the chunk tables, and the reverse must hold. A number on only one side signals an omission somewhere.

Third, continuity of event numbers. Are there no missing numbers from the first event to the last? Numbers may break when events are removed or merged during work.

Fourth, integrity of prior-event references. Does every prior-event number point to an actual event? If an event it points to disappears, the chain of causation breaks.

Fifth, existence of related keywords. Do all people, places, and objects named as an event’s related keywords actually exist? Minute variations in spelling make one thing appear to be another.

Sixth, inter-material references. Do all event numbers cited by the thirty-five materials occur in the event master? In other words, do materials refer to other materials accurately?

Seventh, handling duplicates and integrated files. Where integrated notes and individual notes coexist, as for a married couple, is the source of truth clear? When the same information is split in two places, the authoritative one must be known.

Three Levels of Verification — Automated, Sample, End-to-End

We do not inspect all seven at once, because the nature of checking divides into three. Some checks are mechanically passed or failed without human judgment; others require a human eye to follow a random selection; still others require one complete journey from the beginning to the end of a task. We distinguish these three levels explicitly.

The automated level gives results without human inspection. Are Event Numbers 1 through 49 present without gaps? Does every prior-event reference exist? Is a character’s name spelled the same in every material? A short script handles these. Items 1 through 5 among the seven belong here. It takes almost no time and gives immediate pass or failure. A failure is immediately registered as P0 or P1 and enters the correction queue.

The sample level randomly selects parts of the materials for direct human reading. We open one chunk table to see whether it prematurely contains information unknowable at that point (the spoiler principle), whether an event’s causal chain matches the textual facts, and whether a key-line quotation is exact. Item 6, inter-material reference, is mainly handled here. It takes thirty minutes to an hour; one need not read every material to sense its grain.

The end-to-end level simulates translating one chapter or scene from beginning to end using only the folder’s materials. We select one random chapter and determine translation choices by opening chapter guide → character note → period encyclopedia → sentence tree in sequence. Where no decision can be made is precisely where materials need supplementing. The seventh item, handling duplicates and integrated files, is also naturally checked here, because this simulates how a person newly receiving the folder opens its materials. It takes one or two hours.

Separating these levels matters because passing one does not guarantee passing another. An automated pass can still fail at the sample level in meaning; a sample pass can still reveal friction entering materials at the end-to-end level. Verification has depth only when all three are run separately.

What Was Found in Pride and Prejudice

Applying the seven items to the Pride and Prejudice folder produced about ninety check items. Here are several decisive findings.

Darcy’s younger sister was written “Miss Darcy” in one place and “Georgiana Darcy” in another. Both names occur in the work, but materials must standardize one form; otherwise search treats one person as two. Standardization was needed.

Two background events cited nonexistent prior events. They were remnants of temporary numbering from early work, never cleaned up after formal numbers were assigned. The causal chain required correction.

The folder held notes for three minor figures, while the chunk tables contained no explanatory lines for them. Notes were made first and the tables not yet reinforced. This required reinforcement for traceability.

Which Errors Come First?

Discovered errors do not all carry the same weight. We process them in three grades.

P0 errors directly affect translation accuracy. They are corrected immediately. Inconsistent character-name forms belong here: if a translator sees two forms and understands two people, translation itself goes wrong. Pride and Prejudice had two P0 errors, both corrected at once.

P1 errors affect material traceability. They are corrected before the folder is publicly released or work proceeds to the next stage. Incorrect event references belong here. They do not directly affect translation, but must be fixed to guarantee trustworthiness. Pride and Prejudice had three P1 errors; all were corrected before release.

P2 errors need structural reinforcement. Work remains possible, but the next cycle improves them for completeness. Missing explanatory lines for minor figures belong here. Pride and Prejudice had about eighty-five P2 items, the largest count; they were carried into the next cycle’s work list.

This grading exists because verification would never end if every error were handled with the same time and weight. Priority must be clear: P0 first, then P1, then P2 recorded for the next cycle.

Separating Sources of Truth for Integrated Sheets

The seventh check item—handling duplicates and integrated files—is small but important. Consider one Pride and Prejudice case.

The work contains the Gardiners, Mr. and Mrs. Gardiner. They act almost always together: they take Elizabeth on the Pemberley trip, respond to Lydia’s elopement, and finally witness the wedding.

For such a group, we made two kinds of note: individual notes for Mr. Gardiner and Mrs. Gardiner, and an integrated note for the Gardiners as a couple. But with all three files in one folder, the same information is divided between two places. Mrs. Gardiner’s past residence in Lambton, Derbyshire may appear in both her individual note and the couple’s integrated note.

The question is: which is the source of truth? If the two materials differ slightly, which should be trusted?

Our rule is this: individual notes are the sources of truth. The integrated note is defined as a supplementary material dealing only with the couple’s shared narrative and joint judgment. Its opening therefore carries this guidance:

This document is an integrated sheet for the Gardiners’ relationship and shared narrative role. For detailed information about each person, refer to the individual notes for Mr. Gardiner and Mrs. Gardiner. In translation work, the individual sheets are the sources of truth; this document supplements their paired narrative and joint judgment.

It is a small line, but a critical rule for material reliability. Without a clear source of truth, a worker loses their way when two materials conflict.

When Verification Ends

Once all ninety checks have been handled, one verification report enters the folder’s final layer. It records everything checked, every grade of error found, and how it was treated.

The report is the final seal guaranteeing the folder’s reliability. A person receiving the folder first reads the guide, sees the materials list, and then opens the verification report; at a glance they understand the folder’s state. Knowledge that the materials passed verification becomes the ground of users’ trust.

This is the final stage of the six. From Stage 1 profiling to Stage 6 verification, our work on one work settles into the folder in ordered form.

The Release Threshold

The bare fact that verification has ended does not release a folder. We establish an explicit threshold. Only when all of the following pass can the folder go outside.

Every automated-level check must pass. Event-number continuity must hold; every prior event must point to a real event; character names must use the same form across materials. One automated failure holds release, because failure in an area requiring no human judgment means the most basic integrity has broken.

Every discovered P0 error must be corrected. No item directly affecting translation accuracy may remain unresolved when the folder leaves. P1 errors must either be corrected or have an explicit decision to defer them to the next cycle. They are not merely postponed; the decision not to address them in this cycle is recorded.

The folder guide carries a table of verification status: automated result, sample result, end-to-end result, number of remaining P2 issues, and release pass/fail. Whoever receives the folder can know its state at once from this one table.

The threshold is what gives verification meaning. If one checks but sets no threshold, what passes and what is held remain ambiguous, and that ambiguity erodes material trust. The threshold is the last line dividing release.

What We Have Seen Together over Fourteen Installments

Let us briefly look back over the fourteen and move toward the close.

Across Part 1’s three installments, we established the place of a practice called , drew the six-stage process framework, and saw the classification logic of more than thirty kinds of materials.

Across Part 2’s four installments, we examined the exterior of the Pride and Prejudice work folder; opened one chunk and one group of events directly; saw materials that decide a sentence’s forks in advance; and followed the path by which thirty-five kinds were chosen.

Across Part 3’s three installments, we saw the thirty-five kinds one by one: thirteen skeletal materials for any work, sixteen kinds of flesh giving Pride and Prejudice its color, and six additions unique to it.

Across Part 4’s three installments, we saw character notes in three people: the fullest form of protagonist Elizabeth, the compressed depth of supporting Charlotte, and the one-scene weight of minor Mrs. Reynolds. We also saw supporting files in the folder and how division of labor reduces time.

In this final installment of Part 5 we saw verification, the stage by which material integrity is guaranteed, and now arrive at the reckoning.

Over fourteen installments, we have seen everything made around one book and how it becomes one coherent reference book.

A Greeting to Pride and Prejudice

Here I would like to greet one book.

A book written by an English woman two hundred years ago has lain open on our worktable for fourteen installments: Elizabeth, Darcy, Jane, Bingley, Charlotte, Collins, Wickham, Lydia, the Bennets, the Gardiners, Mrs. Reynolds. We made each one into a note. The tables, carriages, balls, letters, and social seasons of 1813 live in our materials folder.

This book crossed two hundred years to come to us. Because it did, we could make materials for fourteen installments. We give thanks to this one book.

An Honest Admission of Limits

is not a completed system. Each work adds discoveries to the procedure. We should honestly name two or so limits we currently face.

Some stages still require much human work. In particular, at Stage 5, the time needed to decide the structure of one kind of material is hard to automate. Even the same motif has different intensity and expression in different works.

There is also the accumulation of minute notation differences across materials. If workers write one character’s name slightly differently, verification catches it, but it still accumulates as P2 issues. We continue refining tools for standardizing notation.

Standardization for simultaneous multilingual work is not complete either. Creating a materials folder for each of thirty-nine languages is inefficient; for one folder to serve thirty-nine languages at once, its material forms must become more abstract. That is the next task.

These limits do not arise from an immature procedure, but because working with foreign languages essentially requires human judgment. That posture is not replaced by automation.

An Apple Is Not an Apple — Final Return

Let us return to the metaphor of the first installment.

The thirty-five kinds of material, thirty-four character notes, and 156-line event index we made are all the result of applying one sentence to one book:

An apple is not an apple. A pear is not as good as a pear.

To carry the weight with a word when carrying the word itself: that is . Let me say it once more, briefly. When we sit before the next book, we begin again by taking that posture.

To Colleagues Who Work Together

The folder we made ultimately lives through human hands.

Translators, adapters, content planners, dubbing directors, copywriters: colleagues receive our materials and make work with them. Even when materials are exact, the sensibility of the person using them is the final decision-maker. We make materials; our colleagues’ hands make the work live.

I expect some of those colleagues have followed these fourteen installments. Thank you. May these materials live well in the work of those who read them.

Next Lecture Preview

is one axis of our work. There is another: the work of building a keyword library, which we call mejeLIBRARYING. Where the two axes meet, a complete dictionary for one work is made.

The next lecture manuscript will deal with that other axis. If a work’s characters, events, and materials are organized in , all its keywords become a library in mejeLIBRARYING. When the two meet, one book is ready to be carried into 169 countries and 39 languages.

Closing

Thank you for being with us through fourteen installments.

I hope this lecture was not merely the delivery of information, but an invitation to the posture we take before foreign languages: the posture of attending carefully to one work; the posture of recognizing the weight of one word; the posture of not saying an apple is the same as an apple.

May we meet again before the next book.