Latent

Latent is a computational measurement application designed to complement professional review and linguistic research (e.g. Critical Discourse Analysis (CDA) and Legal Informatics). It contributes to ongoing methods to connect qualitative reading with deterministic, quantitative measures of language use, allowing users to map the structural architecture of formal correspondence - such as email, legal, regulatory or client correspondence in an adversarial context. STRUCTURAL LINGUISTIC ANALYSIS Latent does not interpret facts, judge intent or replace professional expertise. Instead, it ignores the written factual payload entirely. The application scans for syntactic choices - such as passivisation, burden-shifting, and epistemic hedging - and translates them into a quantifiable structural matrix (Useful for CDA and Legal Informatics). KEY MEASUREMENTS • DEFENSIVE SHIELDING (Epistemic Modality, Passivisation et al): Measure the density of language used to obscure accountability. • AGGRESSIVE ASSERTION (Deontic Modality, Presupposition, Evaluative Polarity et al): Track structural markers used to force premises and shift burdens. • PROXIMITY DISSONANCE: Detect internal textual friction where highly authoritative demands physically collide with evasive, shielding language. • SYNTHETIC PLAUSIBILITY: Identify mathematically uniform structural signatures of Large Language Models (LLMs). Latent flags text that combines rigid stylistic uniformity with severe institutional shielding (like Epistemic Modality) - a primary indicator of generative AI assistance. THE COMPARATIVE - ASYMMETRY Linguistic strategy cannot be measured in a vacuum. Latent allows a user to cross-examine the documents of two opposing parties to calculate the volume imbalance in their strategic framing. This provides objective telemetry on which party is driving the narrative framework, and which is reacting. DRAFTING CONSISTENCY - COSINE SIMILARITY Human writers naturally fluctuate in their structural approach from document to document. Latent calculates the Cosine Similarity across a party's submissions to measure their strategic consistency. Extreme mathematical uniformity often exposes a rigid institutional playbook or automated drafting, while natural variance indicates a reactive, human-authored response. STRICT DATA PRIVACY Latent is a closed loop analytical application designed for confidential, client-privileged documentation. • ZERO DATA EXTRACTION: All Natural Language Processing (NLP), tokenisation and algorithmic calculations are performed entirely on-device using Apple’s native silicon frameworks. • NO AI TRAINING: No data is transmitted to our servers, nor is it ever harvested to train external Large Language Models (LLMs). • SECURE ICLOUD SYNC: Application data synchronises seamlessly across your own Mac, iPad, and iPhone devices exclusively via your private, encrypted Apple ID container. As Latent continues to evolve, I welcome feedback and critique from professionals and researchers to inform future versions. Note: Latent is a measurement tool, not an automated judge. It is designed to enable professional review with reproducible structural data.