From 38ab434642a6153417abf8a885ab0bf6cf9b5694 Mon Sep 17 00:00:00 2001
From: epi <43392618+epi052@users.noreply.github.com>
Date: Sun, 16 Nov 2025 08:03:58 -0500
Subject: [PATCH] touched up a few minor issues in nlp

---
 src/nlp/document.rs | 5 ++---
 src/nlp/model.rs    | 6 +++++-
 src/nlp/term.rs     | 9 +++++++++
 3 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/src/nlp/document.rs b/src/nlp/document.rs
index c10e9e2..85a163f 100644
--- a/src/nlp/document.rs
+++ b/src/nlp/document.rs
@@ -20,11 +20,10 @@ impl Document {
 
         let processed = preprocess(text);
 
-        document.number_of_terms += processed.len();
-
         for normalized in processed {
             if normalized.len() >= 2 {
-                document.add_term(&normalized)
+                document.add_term(&normalized);
+                document.number_of_terms += 1;
             }
         }
         document
diff --git a/src/nlp/model.rs b/src/nlp/model.rs
index 588d2a3..d01582c 100644
--- a/src/nlp/model.rs
+++ b/src/nlp/model.rs
@@ -73,7 +73,11 @@ impl TfIdf {
                 to_add.push(score);
             }
 
-            let average: f32 = to_add.iter().sum::<f32>() / to_add.len() as f32;
+            let average = if to_add.is_empty() {
+                0.0
+            } else {
+                to_add.iter().sum::<f32>() / to_add.len() as f32
+            };
 
             *metadata.tf_idf_score_mut() = average;
         }
diff --git a/src/nlp/term.rs b/src/nlp/term.rs
index 007b8ce..e37777d 100644
--- a/src/nlp/term.rs
+++ b/src/nlp/term.rs
@@ -22,6 +22,15 @@ impl Term {
 }
 
 /// metadata to be associated with a `Term`
+///
+/// # Design Note
+///
+/// The `count` field represents the number of times a term appeared in a **single document**
+/// and is only meaningful in the per-document context (i.e., within a `Document`).
+///
+/// When `TermMetaData` is stored in the global `TfIdf` model, the `count` field becomes stale
+/// and is not used. Instead, the model relies on `term_frequencies` (which tracks the term
+/// frequency for each document the term appears in) and calculates TF-IDF scores from those.
 #[derive(Debug, Clone, Default)]
 pub(super) struct TermMetaData {
     /// number of times the associated `Term` was seen in a single document