WordCells
Reusable lexical Cells representing words, their stable identities and lexical knowledge. A word may support more than one grammatical or meaning branch, but those branches must remain distinguishable.
Architecture
Cells are the nodes; sentences and semantic fields are the connections
The architecture separates reusable Cell identity from source-specific relationships, while preserving a clear path from every answer back to its SentenceCell evidence.
Updated: July 2026
ActionCells
Reusable action and verb Cells representing actions and verbal forms. They supply persistent action identities without storing sentence-instance backlinks.
SentenceCells
Source-specific Cells that bind token identities, roles, context and compact connections. SentenceCells are the primary learned relationship and the evidence used in retrieval.
Sentence connection
WordCell / ActionCell
↘
SentenceCell — source order, roles, context, evidence
↗
WordCell / ActionCell
The relationship is owned by the SentenceCell. Reusable Cells do not need to point back to every sentence, and facts are not copied into multiple Cell files.
Semantic field connection
Cell fields: key → class → subcategory → category → subdomain → domain
These levels describe a Cell's position within the wider semantic globe. They are not replacement nodes. The Cell remains the node, and its fields provide routes to Cells with related meaning.
Sentence usage determines which meaning and category route is appropriate. A Cell may eventually participate in several field paths when it has multiple meanings or domain uses.
Temporary core activity
The centre of the semantic globe is not a permanent duplicate knowledge store. It is temporary processing state: the selected text, reconstructed discourse anchors, activated Cells, candidate evidence paths and the final answer.
Learning order: sentences first establish relationships and contextual meaning; category fields later extend those relationships across broader knowledge.