An example extraction of concepts, themes, entities, and relationships from a sample of the full Dossier Space corpus, linking each finding to its source and distinguishing explicit statements from inferred concepts. This test used approximately 20 emails, producing a concept list with definitions, supporting excerpts, and a few examples of ambiguous classifications. This required reading message text, beyond the metadata-only scope, leaving attachments excluded.
Nodes are analyst-defined concepts; arrows are labelled relationships supported by the cited records. A connection is not proof of causation. C09, C10 and C11 remain separate rather than forcing a connection unsupported by this sample.
Four inherited records were selected because S23 found substantive text. The new records are therefore an extension sample, not a new probability sample. All 19 lack exact normalized-body duplicates and repeated sampled Message-ID values. That does not establish semantic independence: B05/B10/N05/N06/N07/N08/N11/N15 include quoted or forwarded correspondence.
Shared themes across records are not corroboration of the same event. No thread expansion was performed.
Dossier Concept Simulator is available to signed-in administrators.
Exploring the simulator
- Map: click a concept card for definitions and support. Click a relationship number or use the relationship list to see the typed link and its interpretation limit.
- Evidence: inspect the currently visible excerpts, statement types and source locators.
- Sample simulator: switch to the original four records, restore all 19, or include/exclude records individually. Selected records may still be hidden by the mailbox filter.
- Find a concept: matches concept labels and definitions, not full email contents.
- Mailbox: restrict support to Bailey, Rapp or Pereira.
- Records per concept: require more distinct sampled records before a concept appears.
- Hide forecast / possibility excerpts: removes excerpts explicitly typed as forecasts or possibilities; other types, including questions and uncertainty, remain visible.
- Restore all: resets the sample and filters. Browser refresh also resets them.
Concept definitions and sample support
| Concept | Definition / boundary | Records |
|---|---|---|
| Agreement lifecycle | Drafting, revising, signing or assessing permissions under an agreement; distinguish a proposed action from completion. | B05, N01, N07, N08, N10 |
| Document availability and completeness | Whether required copies or document sets are available; absence of a copy is not absence of an agreement. | B05, B10 |
| Follow-up and handover | Requests, reminders, responsibility assignment and onward document delivery; intentions are not completed transfers. | B05, B10, N01, N05, N10 |
| Identity and communication routing | Resolving recipients or asking which communication route to use; a routing question is not an established rule. | B10, N13 |
| Trade and volume reconciliation | Reviewing trades, deal status, pricing or scheduled versus allocated volumes; underlying records remain unverified. | B02, N05, N11 |
| Regulatory response and filing | Preparing a response or discussing which contracts should be filed; these are participants’ positions, not legal determinations. | N06, N07 |
| Scheduling and coordination | Arranging, postponing or cancelling calls, meetings or activities; tentative timings remain tentative. | N03, N04, N06, N07, N12, N14 |
| Employee learning | Invitations to structured learning; an invitation does not establish attendance or case-specific advice. | N04 |
| Distress expectations | Statements forecasting bankruptcy, default or employment consequences; preserve uncertainty and prediction time. | N15 |
| Third-party news | Externally authored news circulated in a mailbox; receipt does not establish the recipient’s belief or independent corroboration. | N09 |
| Time-exception reporting | A report about exceptions for a reporting period, including explicit reports of none. | B28 |
| Personal and civic communication | Social, leisure or community communications; do not force these into a business-risk interpretation. | N02, N12, N13, N14 |
Relationship evidence
| Relation | Records | Interpretation limit |
|---|---|---|
| Document availability and completeness motivates a request for Follow-up and handover | B05 (E02;E03) | Reported unavailable copies and subsequent reminder. |
| Agreement lifecycle assigns next handling through Follow-up and handover | N01 (E08) | Drafting and forwarding are requested in the same sentence. |
| Trade and volume reconciliation requests supporting records through Follow-up and handover | N05 (E13) | Investigation asks for confirmations or listings. |
| Regulatory response and filing proposes filing of Agreement lifecycle | N07 (E17) | Material deviation is stated as the reason to file; not a legal conclusion. |
| Document availability and completeness is discussed alongside Identity and communication routing | B10 (E04;E06) | Same thread; co-occurrence only, not proof routing caused the document gap. |
| Regulatory response and filing is coordinated through Scheduling and coordination | N06 (E15;E16) | Participants propose a call to discuss the response. |
| Employee learning requires registration and scheduling via Scheduling and coordination | N04 (E11;E12) | Training announcement sets RSVP and session arrangements. |
| Personal and civic communication uses Scheduling and coordination | N12;N14 (E27;E30) | Availability request and conditional rescheduling, not confirmed events. |
| Personal and civic communication asks about Identity and communication routing | N13 (E28;E29) | Proposed consultation raises a routing question. |
Where meaning would be lost by keyword-only coding
- N08: a question about contractual control is answered “No, unfortunately not.” Coding only the question would reverse the meaning.
- B05: a missing executed copy is not proof that the agreement was unexecuted.
- N05: a quoted list says a trade may be booked incorrectly. This remains an unverified possibility.
- N15: bankruptcy and defaults are forecasts by the sender, not verified events. No attempt was made to validate historical outcomes.
- N09: a newsletter is third-party reportage. Its presence does not prove that the recipient read it or shared its claims. It is coded at the news-provenance level; all newsletter subtopics have not been exhaustively mapped.
- N14: timing is tentative and pending approval. B28 explicitly reports no time exceptions.
- N01: the outer sender fields are blank; Wendy is a body signature. Identity has not been resolved beyond that.
Coverage of all 19 records
| ID | Record type | Concepts |
|---|---|---|
| B02 | business correspondence | C05 |
| B05 | business correspondence | C01, C02, C03 |
| B10 | business correspondence | C02, C03, C04 |
| B28 | business correspondence | C11 |
| N01 | business correspondence | C01, C03 |
| N02 | personal correspondence | C12 |
| N03 | corporate announcement | C07 |
| N04 | training announcement | C07, C08 |
| N05 | business correspondence | C03, C05 |
| N06 | business correspondence | C06, C07 |
| N07 | business correspondence | C01, C06, C07 |
| N08 | business correspondence | C01 |
| N09 | third-party newsletter | C10 |
| N10 | business correspondence | C01, C03 |
| N11 | business correspondence | C05 |
| N12 | personal scheduling | C07, C12 |
| N13 | civic correspondence | C04, C12 |
| N14 | personal scheduling announcement | C07, C12 |
| N15 | personal correspondence containing business forecasts | C09 |
Limits and reproduction
This is convenience selection across three small archives, followed by deterministic ranking and recorded body screening. It is neither random sampling of the whole corpus nor a saturation or accuracy study. The text threshold favors longer messages and may exclude meaningful short ones. Personal messages and a newsletter were retained rather than replaced for business relevance. The concepts are not exhaustive.