Update README.md

README.md

@@ -48,6 +48,9 @@ attracted the documents to the dense area are the topic words.
 ![HDBSCAN Document Clusters](images/hdbscan_docs.png)
 
 **4. For each dense area calculate centroid of document vectors in original dimension. (centroid = topic vector)**
+>The red points are outliers and do not get used for calculating the topic vector. The purple points are the documents vectors identified as being part of a dense area are used to calculate the topic vector.
+
+![Topic Vector](images/topic_vector.svg)
 
 **5. Find n-closest word vectors to the resulting topic vector**