diff --git a/README.md b/README.md
index 12542ec..fb50954 100644
--- a/README.md
+++ b/README.md
@@ -11,7 +11,7 @@ import "github.com/keilerkonzept/bitknn"
 
 `bitknn` is a fast [k-nearest neighbors (k-NN)](https://en.wikipedia.org/wiki/K-nearest_neighbors_algorithm) library for `uint64`s, using (bitwise) Hamming distance.
 
-If you need to classify **binary feature vectors that fit into `uint64`s**, this library might be useful. It is fast mainly because we can use cheap bitwise ops (XOR + POPCNT) to calculate distances between `uint64` values. For smaller datasets, the performance of the [neighbor heap](heap.go) is also relevant, and so this part has been tuned here also.
+If you need to classify **binary feature vectors that fit into `uint64`s**, this library might be useful. It is fast mainly because we can use cheap bitwise ops (XOR + POPCNT) to calculate distances between `uint64` values. For smaller datasets, the performance of the [neighbor heap](internal/heap/heap.go) is also relevant, and so this part has been tuned here also.
 
 If your vectors are **longer than 64 bits**, you can [pack](#packing-wide-data) them into `[]uint64` and classify them using the ["wide" model variants](#packing-wide-data).