Class MachineAuthorshipAssessment
- Namespace
- MailFathom.Domain.Emails.Authorship
- Assembly
- MailFathom.Domain.dll
How much one message's own text reads as machine written, and what that reading was reached from.
public sealed record MachineAuthorshipAssessment : IEquatable<MachineAuthorshipAssessment>- Inheritance
-
MachineAuthorshipAssessment
- Implements
- Inherited Members
Remarks
A heuristic estimate and never a measurement. Nothing here proves who or what wrote a message, no model is asked, and no external service is consulted: the whole answer comes from characters in the text the deployment already stored. Every reader of it — a tool description, a page, an agent reading a result — is told so, because a number between zero and one invites being read as a probability and this one is not.
It is an informational value and not a safety indicator. A high likelihood says the text reads as machine written; it does not say the message is unwanted, dishonest, or dangerous, and a great deal of ordinary correspondence is drafted with a text generator by people who mean every word of it. Nothing in this system files, flags, hides, or refuses a message because of it, and no rule reads it — publishing it is the whole of what is done with it, and what to make of it is the reader's.
The likelihood, the band, and the signals are one record because none of them is re-derivable from the others. The signals say what the text carried, the likelihood is what this profile made of that combination, and the band is the reading a caller branches on; a record keeping only the number could not be read back after a weight moved, which is what ProfileRevision exists to make visible.
Properties
- Band
Gets the coarse reading of Likelihood, which is the value a caller branches on.
- Likelihood
Gets how strongly the text read as machine written, from zero to one inclusive.
- NotAssessed
Gets the assessment a message nothing read carries.
- ProfileRevision
Gets the weighting the reading was reached under, or none where nothing read the text.
- Signals
Gets what the text carried, which is what the likelihood was computed from.
- WasAssessed
Gets whether anything read this message's text at all.
Methods
- Assessed(MachineAuthorshipBand, double, MachineAuthorshipSignals, MachineAuthorshipProfileRevision)
Records what one profile made of the signals it read out of a message's text.