kjannette/citation_sentinel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
kjannette
2026-07-28 10:11:28 +00:00
parent 58e85c00f8
commit ab068ba43f
1 changed files with 9 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
README.md
View File

@@ -61,7 +61,7 @@ When a user submits a query, the system enforces groundedness through a multi-l
of the cited chunks. The similarity is calibrated to a 0–1 scale and averaged, producing a single of the cited chunks. The similarity is calibrated to a 0–1 scale and averaged, producing a single
groundedness score that is surfaced to the user with a visual indicator (green/yellow/red). groundedness score that is surfaced to the user with a visual indicator (green/yellow/red).
## Similarity Metrics for Semantic Understanding ## Similarity Metrics for Semantic Understanding - basics
Cosine similarity measures how closely two vectors (representing data like words, images, or preferences) are aligned in a multi-dimensional space by calculating the cosine of the angle between them. Cosine similarity measures how closely two vectors (representing data like words, images, or preferences) are aligned in a multi-dimensional space by calculating the cosine of the angle between them.
@@ -71,6 +71,14 @@ Archetypical cosine scores range from -1 to 1.
0: Vectors are at a 90-degree angle (orthogonal/unrelated). 0: Vectors are at a 90-degree angle (orthogonal/unrelated).
-1: Vectors point in opposite directions. -1: Vectors point in opposite directions.
## Beyond Pure Cosine Similarity
In 2025, V.S. Raghu Parupudi proposed two metrics: Overlap Similarity (OS) and Hyperbolic Tangent Similarity (HTS) as “more robust normalization schemes, [to] capture [holistic semantic similarity more effectively than traditional methods.” Parupudi, V. S. R. (2025). Magnitude Matters: a Superior Class of Similarity Metrics for Holistic Semantic Understanding. arXiv:2509.19323.
Parupudi concluded that “for a wide range of… NLP applications… paraphrase detection, semantic search, and inference - practitioners can achieve… performance improvement by replacing Cosine Similarity with Overlap Similarity or Hyperbolic Tangent Similarity.”
Both OS and HTS attempt to introduce, to varying degrees (and with varying efficacy) 1) relational normalization 2) numerical-stability improvement within a bounded range and 3) outlier suppression.
## Design ## Design
Two-package monorepo: Two-package monorepo: