Terminology
Semantic Terrain is the language of corpus-scale sensemaking. These terms define how CAINC turns institutional records into navigable maps of priorities, risk, entities, boundaries, and change.
Core Concepts
A computed, navigable map of a document corpus's structural geography. A Semantic Terrain is the reading produced when CAINC takes a topological measurement of a full corpus. It reveals how topics cluster, where boundaries form, and how the shape of institutional language changes over time. It is not a search engine, a summary, or a visualization of keywords.
The category of analysis that produces navigable, evidence-backed structure from a bounded document corpus. Where traditional research reads individual documents, corpus intelligence reads the structural relationships between all documents at once. The output is a map of what an institution talks about, how those topics relate, and where the landscape is shifting. Corpus intelligence is what a maintained Semantic Terrain delivers.
The proper name of the SemanticGaps complex epistemic system for taking a topological reading of a document corpus. The name comes from the Welsh word for branch; it is not an acronym. CAINC ingests, embeds, and computes the full body of a named institution into a navigable Semantic Terrain.
A bounded collection of documents from a single institution or domain. A corpus might consist of years of Department of Justice press releases, a decade of FDA enforcement actions, or the full published opinion history of a federal court. Each corpus has a shape. CAINC makes that shape visible and navigable.
A document body that grows as new material is published by the source institution. A live corpus is not static. As the DOJ publishes new press releases, as a regulatory body issues new filings, the corpus expands and the Semantic Terrain updates to reflect its new shape.
A retrieval and generation method in which a model navigates the Semantic Terrain of a document corpus rather than documents alone. TAG works over the shape of the corpus itself, retrieving and reasoning through terrain structures such as peaks, passes, corridors, communities, bundles, and footprints. Documents remain the evidence base, but the model operates on the structural geography of the corpus.
Terrain Features
The dominant topic clusters within a Semantic Terrain. Peaks represent the conceptual high ground of a corpus. In the DOJ terrain, peaks correspond to major enforcement categories: healthcare fraud, national security, public corruption. Each peak is a stable cluster of language that defines a region of institutional focus. Peaks can contain sub-peaks, allowing you to drill down from broad enforcement categories into finer-grained topics within them.
Transition zones where topics overlap, share language, or shift into one another. Passes are the low points between peaks. They reveal where jurisdiction is contested, where enforcement categories converge, or where a single case touches multiple domains. Passes are often where the most interesting structural insights live. Passes are also the gateways between communities. A pass between two peaks that belong to different communities marks the structural boundary where one institutional focus transitions into another.
The paths between connected peaks in a Semantic Terrain. Corridors show how topics relate to one another through shared language and overlapping entities. Following a corridor between two peaks reveals the structural connection between seemingly separate enforcement areas. In CAINC, corridors are navigable. Select two peaks and trace the corridor between them to see the documents, entities, and language that bridge the two topics.
A scoped set of related peaks treated as one structural unit for retrieval and analysis. A bundle is not a single peak and not a whole community. It is a chosen grouping of peaks, often assembled from a route, a comparison, or a focused investigative question. Bundles let you study the combined document field around multiple terrain features at once: the shared entities, the bridging language, the distinctive signals, and the full article set they define together.
Computed boundaries between topic regions within a Semantic Terrain. Contours show where one enforcement focus ends and another begins. They are derived from the actual geometry of the language in the corpus, not from human tagging or manual categorization.
A measure of how dominant and stable a peak is within the terrain. High-persistence peaks represent deeply established institutional focus areas that have remained prominent across the corpus. Low-persistence peaks may represent emerging or peripheral topics. Persistence helps distinguish the signal from the noise in a large corpus.
The process of locating people, organizations, and cases precisely within a Semantic Terrain. Every entity in the corpus occupies a specific position in the terrain. Entity mapping reveals not just that a name appears, but where it sits relative to the surrounding topics, which peaks it is near, and what other entities share its topological neighborhood.
A comprehensive profile of a single entity within a Semantic Terrain. A dossier shows every document the entity appears in, which peaks it is associated with, which other entities it co-occurs with, and how its presence in the corpus has developed over time. Dossiers turn a name into a full structural portrait. Dossiers can also be scoped: when opened from within a specific peak or community, the dossier shows the entity's role within that context, not just its global profile.
The structural imprint of a scoped article set when projected onto the terrain or network. A footprint shows where a selection of documents lands, concentrates, or spreads across the Semantic Terrain. Terrain footprints reveal which peaks and regions carry the scope. Network footprints reveal which entities and relationships define it. A footprint is not just a result list. It is the visible shape of a bounded evidence set inside the larger corpus.
A group of related peaks that form a coherent region within a Semantic Terrain. A community marks a structural area of shared language and topical affinity larger than a single peak but smaller than the full corpus. Each community is named by its dominant content and can be explored as a navigation scope.
The three scales at which communities are detected within a Semantic Terrain: continental, regional, and street. These tiers let you move between broad structural regions and fine-grained local neighborhoods without losing context.
The structural backbone connecting peaks across the terrain. The skeleton shows how the major topics of a corpus are linked, which peaks are neighbors, and where the natural transit paths between them lie. It is the road network of the terrain.
Changes in the shape, peaks, or boundaries of a Semantic Terrain over time as the live corpus grows. Terrain movement requires at least two temporal snapshots of the terrain. A single snapshot shows structure, not movement. When an institution shifts its priorities, when a new topic emerges, or when an existing focus area contracts, terrain movement captures that structural change. Movement is an early hypothesis with evidence, not a verdict.
How You Explore
The primary view of a Semantic Terrain. A topographic visualization showing the density landscape of the corpus, with peaks, contours, and document positions rendered as navigable geography. Click any region to explore the documents and entities within it. Drill into any peak to reveal the finer structure inside.
A view of the entity relationships within a Semantic Terrain. Entities are shown as nodes and their co-occurrences as connections, positioned according to their location in the terrain. The network view reveals which entities appear together, how tightly they cluster, and where they sit relative to the broader topic landscape.
A full-text document browser within CAINC. Search the corpus using natural language or specific terms. Results are ranked by both textual relevance and semantic similarity using vector search, surfacing documents that are structurally related to your query even when they do not share exact keywords.
A dedicated interface for exploring individual entities within the terrain. Select any entity to see its full dossier: the documents it appears in, the peaks it belongs to, the other entities it co-occurs with, and its structural position within the corpus.
Searching within a specific context rather than the full corpus. When exploring a peak, a community, or an entity's documents, scoped retrieval limits search to that context. This surfaces documents that are relevant to your current focus, not just globally similar to a query.
A search method that finds documents by meaning rather than exact keyword match. When you search a Semantic Terrain, vector search identifies documents that are semantically similar to your query, even if they use different words. Combined with full-text search, this surfaces results that keyword-only tools miss entirely.
Services
A Semantic Terrain that is updated on a defined cadence and reflects new documents as they enter the corpus. A living terrain compounds in value over time. Contrast with a static terrain, which is computed once from a fixed corpus snapshot.
Continuous access to a living Semantic Terrain for a named corpus, updated on a defined cadence and used for monitoring, investigation, research, and early structural warning. A terrain subscription is not a one-time report. It is ongoing access to a reading produced by CAINC as the corpus changes. As the institution publishes new material, the terrain updates and subscribers see the movement.
The update frequency of a Semantic Terrain. Cadence depends on the flow of data from the source institution. Some terrains update daily, others weekly, monthly, quarterly, or yearly. The DOJ terrain updates as new press releases are published. Cadence is determined by the natural publication rhythm of the source corpus.
A bespoke Semantic Terrain built to a client's specification from a designated corpus. Organizations that need a specific body of documents mapped can request a custom terrain. The corpus is defined, ingested, and computed into a navigable terrain maintained on an agreed cadence. Custom terrain pricing is handled by direct inquiry.
A scoped analytical engagement using Semantic Terrain data to investigate a specific question or thesis. A deep dive uses the terrain as a starting point, then goes further: tracing entity relationships, analyzing terrain movement over a specific period, or mapping the structural context around a particular case, topic, or enforcement action.
Technical Foundations
The mathematical foundation underlying a Semantic Terrain. A scalar field is a computed density function over the embedding space of the corpus. It assigns a value to every point in the terrain, creating the surface from which peaks, passes, and contours are extracted. The scalar field is what makes a Semantic Terrain a genuine topological structure rather than a visual approximation.
The technical foundations of how Semantic Terrains are computed are documented on our Methodology page. To explore a live Semantic Terrain, visit CAINC. To discuss a terrain for your domain, request a briefing.