Duplicate character record review

A repeated name is a question about records

A cast bible can contain two records for one person, two people with one name, or a title that is used as a name in only one scene. These are different problems. Deleting every repeated string can erase a deliberate namesake. Keeping every row as a separate character can split one person’s history across several entries. Before deciding which spelling to publish, establish what each source record is supposed to describe.

This workbench screens author-supplied fictional records for similar labels, then lets you declare same-person or different-person relationships with reasons. Confirmed same-person links form components. A contradictory different-person relationship produces an inspectable path rather than an automatic compromise. A complete mapping plan also requires an explicit representative for every consistent group containing several records.

The result is a proposal for organizing creative records. It does not edit your manuscript, erase a source, combine biographies or identify a real person. A high score is a character-overlap heuristic, not a probability that two records describe one person. Your scene notes, stable record IDs and decisions provide the context the score cannot supply.

Original fictional scene: two arrivals for Nemi Rook

Nemi Rook held the west floodgate while the market children crossed the little bridge. People called the tender the Reedkeeper when the ceremonial mask was on. After the crossing, a friend leaned from an upstairs window and called Nemi; the tender lifted the mask to answer. The writer had one scene note under Nemi Rook and another under The Reedkeeper. They described the same fictional person in different circumstances.

The ferry brought a second Nemi Rook that afternoon, an instrument repairer carrying a concertina. At the landing, the children began a chorus of thanks to Nemi for holding back the water. The repairer bowed, then explained that no water had been held back: only an instrument had been repaired. Beside the repairer stood the tender, still rubbing tired arms. They introduced themselves, each expecting the other to correct the repeated name. Neither did. The crowd had addressed the wrong person without either person pretending to be someone else.

In an early cast spreadsheet, the writer had joined the two Nemi records by their identical full name. A second link associated the repairer with the Reedkeeper because the title’s source also contained Nemi. Together those links would assign the tender’s actions, the repairer’s instrument work and the masked title to one person. The writer’s explicit note that both people meet at the dock contradicted that interpretation. Removing whichever label looked least similar would not settle the actual identity decision.

The revised scene kept the two namesakes. The friend introduced the chorus as thanks to the Reedkeeper for the bridge crossing, then welcomed the repairer to the market. Neither person had to abandon their name. The tender’s title became useful context; it was not a secret third identity created to repair a spreadsheet. The repairer could join the chorus with the newly working instrument, turning the misunderstanding into cooperation instead of an accusation of impersonation.

The people, mask convention, market and incident are invented creative examples. If the repairer were renamed Ilin Voss, the exact full-name match would disappear and the chorus’s mistake would need another cause. That revision would not automatically prove who the Reedkeeper is. Update the source record and revisit any decision tied to its former contents. A name change can alter a suggestion while leaving the need for actual scene evidence intact.

Prepare a bounded set of source records

Each entry has exactly id, name, aliases, source and note. The ID identifies a record, not an already verified person. Keep the two Nemi rows under different IDs while investigating them. Aliases are supplied labels found in your creative material; the tool does not search the manuscript to discover them. A source is your own location label, such as a scene or session reference, not a verified citation.

The default entries are gate, keeper and passenger. Gate records Nemi Rook at the floodgate. Keeper records The Reedkeeper with the supplied alias Nemi. Passenger records the repairer’s arrival and the authored scene in which the namesakes meet. The two relevant screening suggestions are gate/keeper and gate/passenger. Equal text supplies a useful question in both cases, although their eventual answers differ.

The worksheet accepts 1–24 records, with up to four distinct aliases per record. IDs start with an ASCII letter and use at most 40 letters, digits, underscores or hyphens. Name and alias fields are nonblank single-line Unicode text, limited to 120 UTF-16 units; source is limited to 200 and note to 500. The complete serialized input has its own 750,000-unit limit. Split a larger project into explicitly tracked review scopes rather than dropping inconvenient rows.

Retain the original cast bible. Unsupported extra record columns are rejected so the tool cannot quietly discard your dates, pronouns or production notes. Prepare this naming projection deliberately and keep the correspondence to the full source. A source label helps you find evidence; it does not establish that the evidence is true or that the program has opened it.

What the suggestion score measures

The comparison uses adjacent Unicode code-point pairs, including spaces. Repeated pairs retain their counts. For two forms, it adds the smaller count of each shared pair, doubles that overlap and divides by the total pair counts in both forms. A one-code-point label uses its single token instead. The maximum score across the supplied name/alias combinations becomes the pair’s suggestion score.

For aaaa and aaa, the pair aa occurs three times and twice. Their shared count is two, so the fraction is four fifths, or 80 percent. The strings aab and abb share only one adjacent pair and score 50 percent. An anagram-like inventory is not enough: ab and ba have no matching adjacent pair under this method. These examples describe the implemented arithmetic, not how readers perceive names.

Exact comparison changes nothing. The optional ASCII policy converts A–Z to a–z only. There is no general Unicode case folding, accent removal, transliteration, phonetic comparison or normalization. Similarity in another writing system can behave differently under this deliberately narrow character method. Enter researched or authored alternative labels explicitly instead of assuming the software has inferred them.

The inclusive threshold uses the unrounded fraction; the displayed percentage is rounded to one decimal place. Raising the threshold can remove suggestions without proving those records distinct. The complete plan records the policy and threshold and states that outside pairs were not assessed by this screen. It never converts a chosen cutoff into a guarantee that the whole cast has no duplicate identities.

Declare relationships without asking the score to decide

Open a pair’s controls, choose same, different or defer, and give a reason. Same and different are author declarations about the fictional material. Defer records that you still need to check; it does not count as a completed decision. A blank reason is rejected instead of being replaced with a fabricated explanation.

For gate/keeper, choose same and identify the authored mask-removal scene as the basis. For gate/passenger, choose different and cite the scene where the two people meet. You now have a consistent two-record group plus a separate passenger. The tool has not inspected either scene or certified the author’s interpretation. It has recorded the declared relationships and checked their compatibility.

You can also review a pair outside the lexical suggestions. Use the manual pair controls, which require the same source context and reason. This matters when two labels share no characters but the story explicitly connects them. It also lets you challenge an apparently separate pair without lowering the threshold for every other record in the scope.

Follow a contradiction back to its actual links

Confirmed same-person relationships can be followed as paths. If gate is the same person as keeper, and keeper the same as passenger, the declared relationships put all three in one component. A different-person decision between gate and passenger then contradicts that component. The report lists the negative decision, the positive path and the review IDs along it. No edge is discarded merely because its similarity score is lower.

Different-person relationships do not work in the same transitive way. If A differs from B and B differs from C, nothing in those two statements settles whether A and C are the same. The worksheet leaves their relevant screened pair unresolved. In contrast, a known different relationship between two confirmed groups can cover another cross-group pair with positive paths back to that explicit witness.

Use same for one fictional person within the declared review scope. A shared ancestor, cloned origin or counterpart in another world is a different relationship. Do not encode merely related identities as same to obtain a convenient component. A branching-identity workflow is more suitable when references must distinguish those descendants or counterparts.

Repair a contradiction by returning to the sources and revising the actual declaration that no longer represents your intent. Changing a link to defer still leaves a pending obligation. Removing a stored review is explicit; if its pair remains a screening suggestion, the question becomes unreviewed again. The tool does not reward deletion of evidence with a false completed plan.

Choose what represents a consistent group

A same-person group does not automatically inherit the first row’s name. Select a particular representative record after the relationships are consistent. For the default gate/keeper group, choosing gate keeps Nemi Rook as the representative label in the mapping while preserving the Reedkeeper source row and every supplied label. Passenger remains its own representative.

This choice identifies a reference record for later organization. It does not make that name the only form allowed in dialogue, reveal a masked identity to every observer or discard facts from the other record. Use a separate alias or scene-knowledge workflow when deciding who may know the connection. The review package is private author material and can spoil those connections.

A representative choice retains the exact member records and the confirmed same-person links that define the group. A revised source or changed same-person basis can make it stale. If the group splits or grows, its old choice remains an explicit obsolete obligation until removed or rebuilt. It is not silently transplanted onto a different set of records.

Keep old judgments from becoming new evidence

Each relationship review retains the two source records it was based on. Changing a name, alias, source label or note makes the old review stale and stops its edge from being applied. A review whose endpoint disappears remains visible as an orphan. Inspect and update or remove those judgments; do not assume a source edit has transferred their authority to new material.

Controls check the current form context before recording a decision. If you change a source textarea without rerunning the screen, an old visible button cannot capture its previous expectation against the new record. Recompute first. A stored decision can become current again if its exact source context is restored; this is context matching, not an independently verified historical approval.

Changing the heuristic policy or threshold alone does not rewrite confirmed author facts. It changes which additional pairs need screening. The distinction keeps a visual-comparison preference separate from the writer’s established identity relationships. A complete plan depends on the current scope, current valid reviews and current group choices together.

Carry a plan into production without losing the sources

The private audit can be downloaded while questions remain. The dedicated mapping-plan download appears only after conflicts, stale or deferred reviews, screened open pairs and representative obligations are resolved. It contains the mapping, group members, original records and review context. It does not apply a database merge or manuscript replacement.

Main copying, favorites, comparison and project exports also retain private source context. Two same-label records can be saved independently because their identities are scoped by records and groups. Saving parameters preserves an editable worksheet; a file export provides a portable copy. Inspect the file before sharing it with anyone who should not know a hidden connection.

For a campaign archive, session references replace scene labels without creating another kind of identity logic. Two game masters’ notes can be reviewed by the same author-confirmed relationships and representative choices. Afterward, reconcile actual naming fields or audit manuscript occurrences with the relevant tools. Identity review establishes the proposed correspondence; it leaves the source material and the next creative decision visible.

Is a 100-percent label score proof of one person?

No. It is an exact result under the disclosed character-overlap measure. Deliberate namesakes can score 100. Confirm identity from your authored sources and keep distinct IDs until you have made that decision.

Why does a different-person chain not settle every pair?

A differing from B and B differing from C does not determine A versus C. The tool requires a relevant witness or review; it does not turn all graph connections into identity equivalence.

Can I choose a representative and discard the other rows?

The exported plan preserves every source row. A representative is a reference for organization, not permission to delete biographies or force one public name across all scenes. Apply any later changes separately.

Does a narrower threshold clear the whole cast?

No. It narrows the screened suggestions. The plan records its policy and threshold and discloses outside-pair limits. Existing deferred, stale or orphan reviews still require explicit resolution.