Expand description
Semantic coordinate encoding — Tantivy TF-IDF bridge.
Replaces the SHA-256 hash-based Coordinate5D::encode() with semantically
meaningful coordinates derived from anchor-based term frequency analysis.
Three semantic axes (ported from v2’s anchor embedding + PCA concept):
- x: Logic ↔ Emotion
- y: Micro ↔ Macro
- z: Time ↔ Space
For each axis, two poles of anchor terms are defined. The encoder tokenizes
the input text using Tantivy’s SimpleTokenizer + LowerCaser (consistent
with the search index), computes term frequencies, and projects to [0, 1]
per axis using a smoothed ratio of pole scores.
Structs§
- Semantic
Encoder - Semantic encoder using Tantivy tokenization and anchor-based TF projection.
- Semantic
Scores - Semantic scores for the three content-derived axes.