Update index.html
Browse files- index.html +1941 -18
index.html
CHANGED
|
@@ -1,19 +1,1942 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
</html>
|
|
|
|
| 1 |
+
<!-- Vector Search Simulation By Pejman Ebrahimi -->
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html lang="en">
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-8" />
|
| 6 |
+
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
|
| 7 |
+
<title>Vector Search Methods Comparison</title>
|
| 8 |
+
<style>
|
| 9 |
+
body {
|
| 10 |
+
font-family: "Segoe UI", Tahoma, Geneva, Verdana, sans-serif;
|
| 11 |
+
line-height: 1.6;
|
| 12 |
+
color: #333;
|
| 13 |
+
max-width: 1200px;
|
| 14 |
+
margin: 0 auto;
|
| 15 |
+
padding: 20px;
|
| 16 |
+
background-color: #f5f7fa;
|
| 17 |
+
}
|
| 18 |
+
|
| 19 |
+
h1,
|
| 20 |
+
h2,
|
| 21 |
+
h3 {
|
| 22 |
+
color: #2c3e50;
|
| 23 |
+
}
|
| 24 |
+
|
| 25 |
+
h1 {
|
| 26 |
+
text-align: center;
|
| 27 |
+
margin-bottom: 40px;
|
| 28 |
+
font-size: 2.2em;
|
| 29 |
+
border-bottom: 2px solid #3498db;
|
| 30 |
+
padding-bottom: 10px;
|
| 31 |
+
}
|
| 32 |
+
|
| 33 |
+
.container {
|
| 34 |
+
display: flex;
|
| 35 |
+
flex-wrap: wrap;
|
| 36 |
+
gap: 20px;
|
| 37 |
+
justify-content: center;
|
| 38 |
+
}
|
| 39 |
+
|
| 40 |
+
.search-type {
|
| 41 |
+
flex: 1 1 500px;
|
| 42 |
+
background: white;
|
| 43 |
+
border-radius: 8px;
|
| 44 |
+
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
|
| 45 |
+
margin-bottom: 30px;
|
| 46 |
+
overflow: hidden;
|
| 47 |
+
transition: transform 0.2s;
|
| 48 |
+
}
|
| 49 |
+
|
| 50 |
+
.search-type:hover {
|
| 51 |
+
transform: translateY(-5px);
|
| 52 |
+
}
|
| 53 |
+
|
| 54 |
+
.search-header {
|
| 55 |
+
padding: 15px 20px;
|
| 56 |
+
color: white;
|
| 57 |
+
font-weight: bold;
|
| 58 |
+
font-size: 1.2em;
|
| 59 |
+
}
|
| 60 |
+
|
| 61 |
+
.search-content {
|
| 62 |
+
padding: 20px;
|
| 63 |
+
position: relative;
|
| 64 |
+
}
|
| 65 |
+
|
| 66 |
+
.enn .search-header {
|
| 67 |
+
background-color: #3498db;
|
| 68 |
+
}
|
| 69 |
+
|
| 70 |
+
.ann .search-header {
|
| 71 |
+
background-color: #e74c3c;
|
| 72 |
+
}
|
| 73 |
+
|
| 74 |
+
.semantic .search-header {
|
| 75 |
+
background-color: #2ecc71;
|
| 76 |
+
}
|
| 77 |
+
|
| 78 |
+
.sparse .search-header {
|
| 79 |
+
background-color: #9b59b6;
|
| 80 |
+
}
|
| 81 |
+
|
| 82 |
+
.canvas-container {
|
| 83 |
+
position: relative;
|
| 84 |
+
height: 300px;
|
| 85 |
+
width: 100%;
|
| 86 |
+
background: #f8f9fa;
|
| 87 |
+
border: 1px solid #ddd;
|
| 88 |
+
border-radius: 4px;
|
| 89 |
+
margin-bottom: 15px;
|
| 90 |
+
overflow: hidden;
|
| 91 |
+
}
|
| 92 |
+
|
| 93 |
+
canvas {
|
| 94 |
+
display: block;
|
| 95 |
+
}
|
| 96 |
+
|
| 97 |
+
.controls {
|
| 98 |
+
display: flex;
|
| 99 |
+
justify-content: space-between;
|
| 100 |
+
margin-bottom: 15px;
|
| 101 |
+
flex-wrap: wrap;
|
| 102 |
+
gap: 10px;
|
| 103 |
+
}
|
| 104 |
+
|
| 105 |
+
select,
|
| 106 |
+
button {
|
| 107 |
+
padding: 8px 12px;
|
| 108 |
+
border-radius: 4px;
|
| 109 |
+
border: 1px solid #ccc;
|
| 110 |
+
background: white;
|
| 111 |
+
font-size: 14px;
|
| 112 |
+
}
|
| 113 |
+
|
| 114 |
+
button {
|
| 115 |
+
background: #3498db;
|
| 116 |
+
color: white;
|
| 117 |
+
border: none;
|
| 118 |
+
cursor: pointer;
|
| 119 |
+
transition: background 0.2s;
|
| 120 |
+
}
|
| 121 |
+
|
| 122 |
+
button:hover {
|
| 123 |
+
background: #2980b9;
|
| 124 |
+
}
|
| 125 |
+
|
| 126 |
+
.step-display {
|
| 127 |
+
background: #f0f4f8;
|
| 128 |
+
padding: 15px;
|
| 129 |
+
border-radius: 4px;
|
| 130 |
+
margin-top: 15px;
|
| 131 |
+
font-size: 14px;
|
| 132 |
+
}
|
| 133 |
+
|
| 134 |
+
.step-title {
|
| 135 |
+
font-weight: bold;
|
| 136 |
+
margin-bottom: 8px;
|
| 137 |
+
}
|
| 138 |
+
|
| 139 |
+
.step-description {
|
| 140 |
+
color: #555;
|
| 141 |
+
}
|
| 142 |
+
|
| 143 |
+
ul.features {
|
| 144 |
+
padding-left: 20px;
|
| 145 |
+
}
|
| 146 |
+
|
| 147 |
+
.features li {
|
| 148 |
+
margin-bottom: 5px;
|
| 149 |
+
}
|
| 150 |
+
|
| 151 |
+
.distance-formula {
|
| 152 |
+
font-style: italic;
|
| 153 |
+
background: #f0f0f0;
|
| 154 |
+
padding: 5px;
|
| 155 |
+
border-radius: 4px;
|
| 156 |
+
margin: 5px 0;
|
| 157 |
+
display: inline-block;
|
| 158 |
+
}
|
| 159 |
+
|
| 160 |
+
.tooltip {
|
| 161 |
+
position: absolute;
|
| 162 |
+
background: rgba(0, 0, 0, 0.8);
|
| 163 |
+
color: white;
|
| 164 |
+
padding: 5px 10px;
|
| 165 |
+
border-radius: 4px;
|
| 166 |
+
font-size: 12px;
|
| 167 |
+
z-index: 100;
|
| 168 |
+
pointer-events: none;
|
| 169 |
+
display: none;
|
| 170 |
+
}
|
| 171 |
+
|
| 172 |
+
.legend {
|
| 173 |
+
display: flex;
|
| 174 |
+
flex-wrap: wrap;
|
| 175 |
+
gap: 15px;
|
| 176 |
+
margin-top: 10px;
|
| 177 |
+
}
|
| 178 |
+
|
| 179 |
+
.legend-item {
|
| 180 |
+
display: flex;
|
| 181 |
+
align-items: center;
|
| 182 |
+
font-size: 12px;
|
| 183 |
+
}
|
| 184 |
+
|
| 185 |
+
.legend-color {
|
| 186 |
+
width: 12px;
|
| 187 |
+
height: 12px;
|
| 188 |
+
border-radius: 50%;
|
| 189 |
+
margin-right: 5px;
|
| 190 |
+
}
|
| 191 |
+
|
| 192 |
+
.tabs {
|
| 193 |
+
display: flex;
|
| 194 |
+
margin-bottom: 15px;
|
| 195 |
+
}
|
| 196 |
+
|
| 197 |
+
.tab {
|
| 198 |
+
padding: 8px 15px;
|
| 199 |
+
background: #ddd;
|
| 200 |
+
border: none;
|
| 201 |
+
cursor: pointer;
|
| 202 |
+
border-radius: 4px 4px 0 0;
|
| 203 |
+
margin-right: 2px;
|
| 204 |
+
}
|
| 205 |
+
|
| 206 |
+
.tab.active {
|
| 207 |
+
background: #f0f4f8;
|
| 208 |
+
font-weight: bold;
|
| 209 |
+
}
|
| 210 |
+
|
| 211 |
+
.tab-content {
|
| 212 |
+
display: none;
|
| 213 |
+
background: #f0f4f8;
|
| 214 |
+
padding: 15px;
|
| 215 |
+
border-radius: 0 4px 4px 4px;
|
| 216 |
+
}
|
| 217 |
+
|
| 218 |
+
.tab-content.active {
|
| 219 |
+
display: block;
|
| 220 |
+
}
|
| 221 |
+
|
| 222 |
+
table {
|
| 223 |
+
width: 100%;
|
| 224 |
+
border-collapse: collapse;
|
| 225 |
+
margin: 15px 0;
|
| 226 |
+
}
|
| 227 |
+
|
| 228 |
+
table th,
|
| 229 |
+
table td {
|
| 230 |
+
border: 1px solid #ddd;
|
| 231 |
+
padding: 8px;
|
| 232 |
+
text-align: left;
|
| 233 |
+
}
|
| 234 |
+
|
| 235 |
+
table th {
|
| 236 |
+
background-color: #f0f4f8;
|
| 237 |
+
}
|
| 238 |
+
|
| 239 |
+
tr:nth-child(even) {
|
| 240 |
+
background-color: #f8f9fa;
|
| 241 |
+
}
|
| 242 |
+
|
| 243 |
+
.comparison-table {
|
| 244 |
+
margin-top: 40px;
|
| 245 |
+
}
|
| 246 |
+
|
| 247 |
+
/* Responsive adjustments */
|
| 248 |
+
@media (max-width: 768px) {
|
| 249 |
+
.search-type {
|
| 250 |
+
flex: 1 1 100%;
|
| 251 |
+
}
|
| 252 |
+
|
| 253 |
+
.controls {
|
| 254 |
+
flex-direction: column;
|
| 255 |
+
}
|
| 256 |
+
}
|
| 257 |
+
</style>
|
| 258 |
+
</head>
|
| 259 |
+
<body>
|
| 260 |
+
<h1>Vector Search Methods Comparison Simulation - By Pejman Ebrahimi</h1>
|
| 261 |
+
|
| 262 |
+
<div class="container">
|
| 263 |
+
<!-- ENN Search -->
|
| 264 |
+
<div class="search-type enn">
|
| 265 |
+
<div class="search-header">1. Exact Nearest Neighbor Search (ENN)</div>
|
| 266 |
+
<div class="search-content">
|
| 267 |
+
<p>
|
| 268 |
+
Finds the <strong>exact</strong> closest data points to a query by
|
| 269 |
+
calculating distances to all vectors in the dataset.
|
| 270 |
+
</p>
|
| 271 |
+
|
| 272 |
+
<div class="canvas-container">
|
| 273 |
+
<canvas id="ennCanvas" width="460" height="300"></canvas>
|
| 274 |
+
<div id="ennTooltip" class="tooltip"></div>
|
| 275 |
+
</div>
|
| 276 |
+
|
| 277 |
+
<div class="controls">
|
| 278 |
+
<div>
|
| 279 |
+
<label for="ennDistance">Distance Metric:</label>
|
| 280 |
+
<select id="ennDistance">
|
| 281 |
+
<option value="euclidean">Euclidean (L2)</option>
|
| 282 |
+
<option value="manhattan">Manhattan (L1)</option>
|
| 283 |
+
<option value="cosine">Cosine Similarity</option>
|
| 284 |
+
</select>
|
| 285 |
+
</div>
|
| 286 |
+
|
| 287 |
+
<div>
|
| 288 |
+
<label for="ennStep">Step:</label>
|
| 289 |
+
<select id="ennStep">
|
| 290 |
+
<option value="0">0. Data points</option>
|
| 291 |
+
<option value="1">1. Calculate all distances</option>
|
| 292 |
+
<option value="2">2. Sort by distance</option>
|
| 293 |
+
<option value="3">3. Return nearest neighbors</option>
|
| 294 |
+
</select>
|
| 295 |
+
</div>
|
| 296 |
+
</div>
|
| 297 |
+
|
| 298 |
+
<div class="step-display">
|
| 299 |
+
<div class="step-title" id="ennStepTitle">Step 0: Data points</div>
|
| 300 |
+
<div class="step-description" id="ennStepDesc">
|
| 301 |
+
Initial dataset with vectors in feature space. The query point
|
| 302 |
+
(red) will be compared against all data points.
|
| 303 |
+
</div>
|
| 304 |
+
</div>
|
| 305 |
+
|
| 306 |
+
<div class="legend">
|
| 307 |
+
<div class="legend-item">
|
| 308 |
+
<div class="legend-color" style="background: #3498db"></div>
|
| 309 |
+
<span>Dataset Points</span>
|
| 310 |
+
</div>
|
| 311 |
+
<div class="legend-item">
|
| 312 |
+
<div class="legend-color" style="background: #e74c3c"></div>
|
| 313 |
+
<span>Query Point</span>
|
| 314 |
+
</div>
|
| 315 |
+
<div class="legend-item">
|
| 316 |
+
<div class="legend-color" style="background: #2ecc71"></div>
|
| 317 |
+
<span>Nearest Neighbor</span>
|
| 318 |
+
</div>
|
| 319 |
+
</div>
|
| 320 |
+
|
| 321 |
+
<h3>Key Features:</h3>
|
| 322 |
+
<ul class="features">
|
| 323 |
+
<li>100% accuracy - finds the true nearest neighbors</li>
|
| 324 |
+
<li>
|
| 325 |
+
Computationally expensive for large datasets (O(n) complexity)
|
| 326 |
+
</li>
|
| 327 |
+
<li>
|
| 328 |
+
Becomes inefficient in high dimensions (curse of dimensionality)
|
| 329 |
+
</li>
|
| 330 |
+
<li>
|
| 331 |
+
Simple implementation - just calculate all distances and sort
|
| 332 |
+
</li>
|
| 333 |
+
</ul>
|
| 334 |
+
</div>
|
| 335 |
+
</div>
|
| 336 |
+
|
| 337 |
+
<!-- ANN Search -->
|
| 338 |
+
<div class="search-type ann">
|
| 339 |
+
<div class="search-header">
|
| 340 |
+
2. Approximate Nearest Neighbor Search (ANN)
|
| 341 |
+
</div>
|
| 342 |
+
<div class="search-content">
|
| 343 |
+
<p>
|
| 344 |
+
Sacrifices perfect accuracy for <strong>speed</strong> by using
|
| 345 |
+
efficient data structures to approximate nearest neighbors.
|
| 346 |
+
</p>
|
| 347 |
+
|
| 348 |
+
<div class="canvas-container">
|
| 349 |
+
<canvas id="annCanvas" width="460" height="300"></canvas>
|
| 350 |
+
<div id="annTooltip" class="tooltip"></div>
|
| 351 |
+
</div>
|
| 352 |
+
|
| 353 |
+
<div class="controls">
|
| 354 |
+
<div>
|
| 355 |
+
<label for="annAlgorithm">Algorithm:</label>
|
| 356 |
+
<select id="annAlgorithm">
|
| 357 |
+
<option value="hnsw">Hierarchical NSW</option>
|
| 358 |
+
<option value="pq">Product Quantization</option>
|
| 359 |
+
<option value="lsh">Locality-Sensitive Hashing</option>
|
| 360 |
+
</select>
|
| 361 |
+
</div>
|
| 362 |
+
|
| 363 |
+
<div>
|
| 364 |
+
<label for="annStep">Step:</label>
|
| 365 |
+
<select id="annStep">
|
| 366 |
+
<option value="0">0. Indexed structure</option>
|
| 367 |
+
<option value="1">1. Navigate to region</option>
|
| 368 |
+
<option value="2">2. Local search</option>
|
| 369 |
+
<option value="3">3. Return approximate NN</option>
|
| 370 |
+
</select>
|
| 371 |
+
</div>
|
| 372 |
+
</div>
|
| 373 |
+
|
| 374 |
+
<div class="step-display">
|
| 375 |
+
<div class="step-title" id="annStepTitle">
|
| 376 |
+
Step 0: Indexed structure
|
| 377 |
+
</div>
|
| 378 |
+
<div class="step-description" id="annStepDesc">
|
| 379 |
+
Data is pre-organized into efficient lookup structures that
|
| 380 |
+
cluster or partition the vector space for faster searching.
|
| 381 |
+
</div>
|
| 382 |
+
</div>
|
| 383 |
+
|
| 384 |
+
<div class="legend">
|
| 385 |
+
<div class="legend-item">
|
| 386 |
+
<div class="legend-color" style="background: #3498db"></div>
|
| 387 |
+
<span>Dataset Points</span>
|
| 388 |
+
</div>
|
| 389 |
+
<div class="legend-item">
|
| 390 |
+
<div class="legend-color" style="background: #e74c3c"></div>
|
| 391 |
+
<span>Query Point</span>
|
| 392 |
+
</div>
|
| 393 |
+
<div class="legend-item">
|
| 394 |
+
<div class="legend-color" style="background: #f39c12"></div>
|
| 395 |
+
<span>Search Region</span>
|
| 396 |
+
</div>
|
| 397 |
+
<div class="legend-item">
|
| 398 |
+
<div class="legend-color" style="background: #2ecc71"></div>
|
| 399 |
+
<span>Returned Neighbors</span>
|
| 400 |
+
</div>
|
| 401 |
+
</div>
|
| 402 |
+
|
| 403 |
+
<h3>Key Features:</h3>
|
| 404 |
+
<ul class="features">
|
| 405 |
+
<li>
|
| 406 |
+
Much faster than ENN for large datasets (sub-linear time
|
| 407 |
+
complexity)
|
| 408 |
+
</li>
|
| 409 |
+
<li>Trades accuracy for speed (95-99% accurate typically)</li>
|
| 410 |
+
<li>Requires pre-processing to build index structures</li>
|
| 411 |
+
<li>Various algorithms optimized for different use cases</li>
|
| 412 |
+
</ul>
|
| 413 |
+
</div>
|
| 414 |
+
</div>
|
| 415 |
+
|
| 416 |
+
<!-- Semantic Search -->
|
| 417 |
+
<div class="search-type semantic">
|
| 418 |
+
<div class="search-header">3. Semantic Search</div>
|
| 419 |
+
<div class="search-content">
|
| 420 |
+
<p>
|
| 421 |
+
Uses <strong>meaning</strong> of content rather than keywords by
|
| 422 |
+
searching through dense embedding vectors that capture semantic
|
| 423 |
+
relationships.
|
| 424 |
+
</p>
|
| 425 |
+
|
| 426 |
+
<div class="canvas-container">
|
| 427 |
+
<canvas id="semanticCanvas" width="460" height="300"></canvas>
|
| 428 |
+
<div id="semanticTooltip" class="tooltip"></div>
|
| 429 |
+
</div>
|
| 430 |
+
|
| 431 |
+
<div class="controls">
|
| 432 |
+
<div>
|
| 433 |
+
<label for="semanticModel">Embedding Model:</label>
|
| 434 |
+
<select id="semanticModel">
|
| 435 |
+
<option value="bert">BERT</option>
|
| 436 |
+
<option value="use">Universal Sentence Encoder</option>
|
| 437 |
+
<option value="custom">Domain-Specific</option>
|
| 438 |
+
</select>
|
| 439 |
+
</div>
|
| 440 |
+
|
| 441 |
+
<div>
|
| 442 |
+
<label for="semanticStep">Step:</label>
|
| 443 |
+
<select id="semanticStep">
|
| 444 |
+
<option value="0">0. Text documents</option>
|
| 445 |
+
<option value="1">1. Generate embeddings</option>
|
| 446 |
+
<option value="2">2. Vector similarity search</option>
|
| 447 |
+
<option value="3">3. Return relevant results</option>
|
| 448 |
+
</select>
|
| 449 |
+
</div>
|
| 450 |
+
</div>
|
| 451 |
+
|
| 452 |
+
<div class="step-display">
|
| 453 |
+
<div class="step-title" id="semanticStepTitle">
|
| 454 |
+
Step 0: Text documents
|
| 455 |
+
</div>
|
| 456 |
+
<div class="step-description" id="semanticStepDesc">
|
| 457 |
+
Starting with raw text documents or queries before encoding into
|
| 458 |
+
vector space.
|
| 459 |
+
</div>
|
| 460 |
+
</div>
|
| 461 |
+
|
| 462 |
+
<div class="legend">
|
| 463 |
+
<div class="legend-item">
|
| 464 |
+
<div class="legend-color" style="background: #3498db"></div>
|
| 465 |
+
<span>Document Embeddings</span>
|
| 466 |
+
</div>
|
| 467 |
+
<div class="legend-item">
|
| 468 |
+
<div class="legend-color" style="background: #e74c3c"></div>
|
| 469 |
+
<span>Query Embedding</span>
|
| 470 |
+
</div>
|
| 471 |
+
<div class="legend-item">
|
| 472 |
+
<div class="legend-color" style="background: #2ecc71"></div>
|
| 473 |
+
<span>Semantic Matches</span>
|
| 474 |
+
</div>
|
| 475 |
+
</div>
|
| 476 |
+
|
| 477 |
+
<h3>Key Features:</h3>
|
| 478 |
+
<ul class="features">
|
| 479 |
+
<li>Understands meaning beyond exact keyword matches</li>
|
| 480 |
+
<li>
|
| 481 |
+
Uses dense vector embeddings (typically 768-1536 dimensions)
|
| 482 |
+
</li>
|
| 483 |
+
<li>Trained on large text corpora to capture language patterns</li>
|
| 484 |
+
<li>
|
| 485 |
+
Effective for natural language, images, and multimodal content
|
| 486 |
+
</li>
|
| 487 |
+
<li>Usually implemented with ANN algorithms for efficiency</li>
|
| 488 |
+
</ul>
|
| 489 |
+
</div>
|
| 490 |
+
</div>
|
| 491 |
+
|
| 492 |
+
<!-- Sparse Vector Search -->
|
| 493 |
+
<div class="search-type sparse">
|
| 494 |
+
<div class="search-header">4. Sparse Vector Search</div>
|
| 495 |
+
<div class="search-content">
|
| 496 |
+
<p>
|
| 497 |
+
Uses <strong>high-dimensional sparse vectors</strong> where most
|
| 498 |
+
elements are zero, optimized for keyword and token matching.
|
| 499 |
+
</p>
|
| 500 |
+
|
| 501 |
+
<div class="canvas-container">
|
| 502 |
+
<canvas id="sparseCanvas" width="460" height="300"></canvas>
|
| 503 |
+
<div id="sparseTooltip" class="tooltip"></div>
|
| 504 |
+
</div>
|
| 505 |
+
|
| 506 |
+
<div class="controls">
|
| 507 |
+
<div>
|
| 508 |
+
<label for="sparseModel">Representation:</label>
|
| 509 |
+
<select id="sparseModel">
|
| 510 |
+
<option value="tfidf">TF-IDF</option>
|
| 511 |
+
<option value="bm25">BM25</option>
|
| 512 |
+
<option value="hybrid">Hybrid (Sparse+Dense)</option>
|
| 513 |
+
</select>
|
| 514 |
+
</div>
|
| 515 |
+
|
| 516 |
+
<div>
|
| 517 |
+
<label for="sparseStep">Step:</label>
|
| 518 |
+
<select id="sparseStep">
|
| 519 |
+
<option value="0">0. Tokenized content</option>
|
| 520 |
+
<option value="1">1. Create sparse vectors</option>
|
| 521 |
+
<option value="2">2. Inverted index search</option>
|
| 522 |
+
<option value="3">3. Return matches</option>
|
| 523 |
+
</select>
|
| 524 |
+
</div>
|
| 525 |
+
</div>
|
| 526 |
+
|
| 527 |
+
<div class="step-display">
|
| 528 |
+
<div class="step-title" id="sparseStepTitle">
|
| 529 |
+
Step 0: Tokenized content
|
| 530 |
+
</div>
|
| 531 |
+
<div class="step-description" id="sparseStepDesc">
|
| 532 |
+
Documents broken down into tokens (words/terms) before converting
|
| 533 |
+
to sparse vector representation.
|
| 534 |
+
</div>
|
| 535 |
+
</div>
|
| 536 |
+
|
| 537 |
+
<div class="legend">
|
| 538 |
+
<div class="legend-item">
|
| 539 |
+
<div class="legend-color" style="background: #3498db"></div>
|
| 540 |
+
<span>Vocabulary Dimensions</span>
|
| 541 |
+
</div>
|
| 542 |
+
<div class="legend-item">
|
| 543 |
+
<div class="legend-color" style="background: #e74c3c"></div>
|
| 544 |
+
<span>Query Terms</span>
|
| 545 |
+
</div>
|
| 546 |
+
<div class="legend-item">
|
| 547 |
+
<div class="legend-color" style="background: #2ecc71"></div>
|
| 548 |
+
<span>Matching Terms</span>
|
| 549 |
+
</div>
|
| 550 |
+
</div>
|
| 551 |
+
|
| 552 |
+
<h3>Key Features:</h3>
|
| 553 |
+
<ul class="features">
|
| 554 |
+
<li>Efficient for exact matching and keyword search</li>
|
| 555 |
+
<li>Very high dimensionality (vocabulary size) but mostly zeros</li>
|
| 556 |
+
<li>Uses specialized inverted index for quick lookup</li>
|
| 557 |
+
<li>Good for precision when exact matches are required</li>
|
| 558 |
+
<li>Often combined with semantic search for hybrid approaches</li>
|
| 559 |
+
</ul>
|
| 560 |
+
</div>
|
| 561 |
+
</div>
|
| 562 |
+
</div>
|
| 563 |
+
|
| 564 |
+
<div class="comparison-table">
|
| 565 |
+
<h2>Comparison of Vector Search Methods</h2>
|
| 566 |
+
<table>
|
| 567 |
+
<thead>
|
| 568 |
+
<tr>
|
| 569 |
+
<th>Feature</th>
|
| 570 |
+
<th>Exact NN (ENN)</th>
|
| 571 |
+
<th>Approximate NN (ANN)</th>
|
| 572 |
+
<th>Semantic Search</th>
|
| 573 |
+
<th>Sparse Vector Search</th>
|
| 574 |
+
</tr>
|
| 575 |
+
</thead>
|
| 576 |
+
<tbody>
|
| 577 |
+
<tr>
|
| 578 |
+
<td>Accuracy</td>
|
| 579 |
+
<td>100% exact</td>
|
| 580 |
+
<td>High (95-99%)</td>
|
| 581 |
+
<td>Context dependent</td>
|
| 582 |
+
<td>High for exact matches</td>
|
| 583 |
+
</tr>
|
| 584 |
+
<tr>
|
| 585 |
+
<td>Speed</td>
|
| 586 |
+
<td>Slow (O(n))</td>
|
| 587 |
+
<td>Fast (sub-linear)</td>
|
| 588 |
+
<td>Moderate to fast</td>
|
| 589 |
+
<td>Very fast for keywords</td>
|
| 590 |
+
</tr>
|
| 591 |
+
<tr>
|
| 592 |
+
<td>Scalability</td>
|
| 593 |
+
<td>Poor</td>
|
| 594 |
+
<td>Good</td>
|
| 595 |
+
<td>Good with ANN</td>
|
| 596 |
+
<td>Excellent</td>
|
| 597 |
+
</tr>
|
| 598 |
+
<tr>
|
| 599 |
+
<td>Vector Type</td>
|
| 600 |
+
<td>Dense or Sparse</td>
|
| 601 |
+
<td>Usually Dense</td>
|
| 602 |
+
<td>Dense</td>
|
| 603 |
+
<td>Sparse</td>
|
| 604 |
+
</tr>
|
| 605 |
+
<tr>
|
| 606 |
+
<td>Use Cases</td>
|
| 607 |
+
<td>Small datasets, high precision required</td>
|
| 608 |
+
<td>Large-scale vector search, recommenders</td>
|
| 609 |
+
<td>NLP, content discovery, similar item search</td>
|
| 610 |
+
<td>Search engines, document retrieval</td>
|
| 611 |
+
</tr>
|
| 612 |
+
<tr>
|
| 613 |
+
<td>Common Metrics</td>
|
| 614 |
+
<td>Euclidean, Manhattan, Cosine</td>
|
| 615 |
+
<td>Euclidean, Inner Product, Cosine</td>
|
| 616 |
+
<td>Cosine, Dot Product</td>
|
| 617 |
+
<td>Jaccard, BM25, TF-IDF</td>
|
| 618 |
+
</tr>
|
| 619 |
+
<tr>
|
| 620 |
+
<td>Dimensions</td>
|
| 621 |
+
<td>Any</td>
|
| 622 |
+
<td>Moderate to high</td>
|
| 623 |
+
<td>High (768-1536 typical)</td>
|
| 624 |
+
<td>Very high (vocabulary size)</td>
|
| 625 |
+
</tr>
|
| 626 |
+
<tr>
|
| 627 |
+
<td>Example Tools</td>
|
| 628 |
+
<td>SciPy, NumPy</td>
|
| 629 |
+
<td>FAISS, Annoy, HNSW</td>
|
| 630 |
+
<td>Pinecone, Weaviate, Milvus</td>
|
| 631 |
+
<td>Elasticsearch, Lucene</td>
|
| 632 |
+
</tr>
|
| 633 |
+
</tbody>
|
| 634 |
+
</table>
|
| 635 |
+
</div>
|
| 636 |
+
|
| 637 |
+
<script>
|
| 638 |
+
// Common data and utility functions
|
| 639 |
+
const dataPoints = [
|
| 640 |
+
{ id: 1, x: 80, y: 70, label: "P1" },
|
| 641 |
+
{ id: 2, x: 160, y: 120, label: "P2" },
|
| 642 |
+
{ id: 3, x: 240, y: 60, label: "P3" },
|
| 643 |
+
{ id: 4, x: 300, y: 180, label: "P4" },
|
| 644 |
+
{ id: 5, x: 400, y: 90, label: "P5" },
|
| 645 |
+
{ id: 6, x: 180, y: 220, label: "P6" },
|
| 646 |
+
{ id: 7, x: 320, y: 260, label: "P7" },
|
| 647 |
+
{ id: 8, x: 370, y: 150, label: "P8" },
|
| 648 |
+
{ id: 9, x: 130, y: 180, label: "P9" },
|
| 649 |
+
];
|
| 650 |
+
|
| 651 |
+
const queryPoint = { x: 220, y: 140, label: "Q" };
|
| 652 |
+
|
| 653 |
+
// Semantic search "documents"
|
| 654 |
+
const semanticDocs = [
|
| 655 |
+
{ id: 1, text: "How to train a dog", embedding: [0.2, 0.7] },
|
| 656 |
+
{ id: 2, text: "Dog training techniques", embedding: [0.25, 0.65] },
|
| 657 |
+
{ id: 3, text: "Cat behavior explained", embedding: [0.7, 0.3] },
|
| 658 |
+
{ id: 4, text: "Pet care for beginners", embedding: [0.4, 0.5] },
|
| 659 |
+
{ id: 5, text: "Feline health issues", embedding: [0.8, 0.2] },
|
| 660 |
+
{ id: 6, text: "Training puppies at home", embedding: [0.15, 0.75] },
|
| 661 |
+
{ id: 7, text: "Bird watching guide", embedding: [0.9, 0.7] },
|
| 662 |
+
{ id: 8, text: "Exotic pet ownership", embedding: [0.6, 0.8] },
|
| 663 |
+
{ id: 9, text: "Dog breeds comparison", embedding: [0.3, 0.6] },
|
| 664 |
+
];
|
| 665 |
+
|
| 666 |
+
const semanticQuery = {
|
| 667 |
+
text: "How to train my puppy",
|
| 668 |
+
embedding: [0.2, 0.8],
|
| 669 |
+
};
|
| 670 |
+
|
| 671 |
+
// Sparse vector "documents"
|
| 672 |
+
const vocabulary = [
|
| 673 |
+
"dog",
|
| 674 |
+
"cat",
|
| 675 |
+
"train",
|
| 676 |
+
"pet",
|
| 677 |
+
"health",
|
| 678 |
+
"food",
|
| 679 |
+
"guide",
|
| 680 |
+
"home",
|
| 681 |
+
"behavior",
|
| 682 |
+
"puppy",
|
| 683 |
+
];
|
| 684 |
+
|
| 685 |
+
const sparseVectors = [
|
| 686 |
+
{
|
| 687 |
+
id: 1,
|
| 688 |
+
text: "Dog training guide",
|
| 689 |
+
vector: [0.8, 0, 0.7, 0.1, 0, 0, 0.3, 0, 0, 0],
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
id: 2,
|
| 693 |
+
text: "Cat health and food",
|
| 694 |
+
vector: [0, 0.9, 0, 0.2, 0.7, 0.6, 0, 0, 0, 0],
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
id: 3,
|
| 698 |
+
text: "Puppy behavior at home",
|
| 699 |
+
vector: [0.3, 0, 0, 0, 0, 0, 0, 0.7, 0.8, 0.9],
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
id: 4,
|
| 703 |
+
text: "Pet food guide",
|
| 704 |
+
vector: [0, 0, 0, 0.7, 0, 0.8, 0.6, 0, 0, 0],
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
id: 5,
|
| 708 |
+
text: "Cat and dog behavior",
|
| 709 |
+
vector: [0.5, 0.5, 0, 0, 0, 0, 0, 0, 0.9, 0],
|
| 710 |
+
},
|
| 711 |
+
{
|
| 712 |
+
id: 6,
|
| 713 |
+
text: "Training your puppy",
|
| 714 |
+
vector: [0, 0, 0.8, 0, 0, 0, 0, 0, 0, 0.8],
|
| 715 |
+
},
|
| 716 |
+
];
|
| 717 |
+
|
| 718 |
+
const sparseQuery = {
|
| 719 |
+
text: "dog training puppies",
|
| 720 |
+
vector: [0.6, 0, 0.7, 0, 0, 0, 0, 0, 0, 0.5],
|
| 721 |
+
};
|
| 722 |
+
|
| 723 |
+
// Distance functions
|
| 724 |
+
function euclideanDistance(p1, p2) {
|
| 725 |
+
return Math.sqrt(Math.pow(p1.x - p2.x, 2) + Math.pow(p1.y - p2.y, 2));
|
| 726 |
+
}
|
| 727 |
+
|
| 728 |
+
function manhattanDistance(p1, p2) {
|
| 729 |
+
return Math.abs(p1.x - p2.x) + Math.abs(p1.y - p2.y);
|
| 730 |
+
}
|
| 731 |
+
|
| 732 |
+
function cosineDistance(p1, p2) {
|
| 733 |
+
// Convert to vectors from origin
|
| 734 |
+
const dotProduct = p1.x * p2.x + p1.y * p2.y;
|
| 735 |
+
const mag1 = Math.sqrt(p1.x * p1.x + p1.y * p1.y);
|
| 736 |
+
const mag2 = Math.sqrt(p2.x * p2.x + p2.y * p2.y);
|
| 737 |
+
return 1 - dotProduct / (mag1 * mag2);
|
| 738 |
+
}
|
| 739 |
+
|
| 740 |
+
function cosineSimilarity(v1, v2) {
|
| 741 |
+
let dotProduct = 0;
|
| 742 |
+
let mag1 = 0;
|
| 743 |
+
let mag2 = 0;
|
| 744 |
+
|
| 745 |
+
for (let i = 0; i < v1.length; i++) {
|
| 746 |
+
dotProduct += v1[i] * v2[i];
|
| 747 |
+
mag1 += v1[i] * v1[i];
|
| 748 |
+
mag2 += v2[i] * v2[i];
|
| 749 |
+
}
|
| 750 |
+
|
| 751 |
+
mag1 = Math.sqrt(mag1);
|
| 752 |
+
mag2 = Math.sqrt(mag2);
|
| 753 |
+
|
| 754 |
+
return dotProduct / (mag1 * mag2);
|
| 755 |
+
}
|
| 756 |
+
|
| 757 |
+
// ENN Canvas Setup
|
| 758 |
+
const ennCanvas = document.getElementById("ennCanvas");
|
| 759 |
+
const ennCtx = ennCanvas.getContext("2d");
|
| 760 |
+
const ennDistanceSelect = document.getElementById("ennDistance");
|
| 761 |
+
const ennStepSelect = document.getElementById("ennStep");
|
| 762 |
+
const ennStepTitle = document.getElementById("ennStepTitle");
|
| 763 |
+
const ennStepDesc = document.getElementById("ennStepDesc");
|
| 764 |
+
const ennTooltip = document.getElementById("ennTooltip");
|
| 765 |
+
|
| 766 |
+
// ANN Canvas Setup
|
| 767 |
+
const annCanvas = document.getElementById("annCanvas");
|
| 768 |
+
const annCtx = annCanvas.getContext("2d");
|
| 769 |
+
const annAlgorithmSelect = document.getElementById("annAlgorithm");
|
| 770 |
+
const annStepSelect = document.getElementById("annStep");
|
| 771 |
+
const annStepTitle = document.getElementById("annStepTitle");
|
| 772 |
+
const annStepDesc = document.getElementById("annStepDesc");
|
| 773 |
+
const annTooltip = document.getElementById("annTooltip");
|
| 774 |
+
|
| 775 |
+
// Semantic Canvas Setup
|
| 776 |
+
const semanticCanvas = document.getElementById("semanticCanvas");
|
| 777 |
+
const semanticCtx = semanticCanvas.getContext("2d");
|
| 778 |
+
const semanticModelSelect = document.getElementById("semanticModel");
|
| 779 |
+
const semanticStepSelect = document.getElementById("semanticStep");
|
| 780 |
+
const semanticStepTitle = document.getElementById("semanticStepTitle");
|
| 781 |
+
const semanticStepDesc = document.getElementById("semanticStepDesc");
|
| 782 |
+
const semanticTooltip = document.getElementById("semanticTooltip");
|
| 783 |
+
|
| 784 |
+
// Sparse Canvas Setup
|
| 785 |
+
const sparseCanvas = document.getElementById("sparseCanvas");
|
| 786 |
+
const sparseCtx = sparseCanvas.getContext("2d");
|
| 787 |
+
const sparseModelSelect = document.getElementById("sparseModel");
|
| 788 |
+
const sparseStepSelect = document.getElementById("sparseStep");
|
| 789 |
+
const sparseStepTitle = document.getElementById("sparseStepTitle");
|
| 790 |
+
const sparseStepDesc = document.getElementById("sparseStepDesc");
|
| 791 |
+
const sparseTooltip = document.getElementById("sparseTooltip");
|
| 792 |
+
|
| 793 |
+
// Event listeners for ENN
|
| 794 |
+
ennDistanceSelect.addEventListener("change", renderENNSearch);
|
| 795 |
+
ennStepSelect.addEventListener("change", renderENNSearch);
|
| 796 |
+
|
| 797 |
+
// Event listeners for ANN
|
| 798 |
+
annAlgorithmSelect.addEventListener("change", renderANNSearch);
|
| 799 |
+
annStepSelect.addEventListener("change", renderANNSearch);
|
| 800 |
+
|
| 801 |
+
// Event listeners for Semantic Search
|
| 802 |
+
semanticModelSelect.addEventListener("change", renderSemanticSearch);
|
| 803 |
+
semanticStepSelect.addEventListener("change", renderSemanticSearch);
|
| 804 |
+
|
| 805 |
+
// Event listeners for Sparse Vector Search
|
| 806 |
+
sparseModelSelect.addEventListener("change", renderSparseSearch);
|
| 807 |
+
sparseStepSelect.addEventListener("change", renderSparseSearch);
|
| 808 |
+
|
| 809 |
+
// Draw all visualizations initially
|
| 810 |
+
renderENNSearch();
|
| 811 |
+
renderANNSearch();
|
| 812 |
+
renderSemanticSearch();
|
| 813 |
+
renderSparseSearch();
|
| 814 |
+
|
| 815 |
+
// ENN Search visualization
|
| 816 |
+
function renderENNSearch() {
|
| 817 |
+
const distanceMetric = ennDistanceSelect.value;
|
| 818 |
+
const step = parseInt(ennStepSelect.value);
|
| 819 |
+
|
| 820 |
+
// Clear canvas
|
| 821 |
+
ennCtx.clearRect(0, 0, ennCanvas.width, ennCanvas.height);
|
| 822 |
+
|
| 823 |
+
// Draw grid
|
| 824 |
+
drawGrid(ennCtx);
|
| 825 |
+
|
| 826 |
+
// Calculate distances based on selected metric
|
| 827 |
+
let distances = dataPoints.map((point) => {
|
| 828 |
+
let dist;
|
| 829 |
+
if (distanceMetric === "euclidean") {
|
| 830 |
+
dist = euclideanDistance(point, queryPoint);
|
| 831 |
+
} else if (distanceMetric === "manhattan") {
|
| 832 |
+
dist = manhattanDistance(point, queryPoint);
|
| 833 |
+
} else if (distanceMetric === "cosine") {
|
| 834 |
+
dist = cosineDistance(point, queryPoint);
|
| 835 |
+
}
|
| 836 |
+
return { ...point, distance: dist };
|
| 837 |
+
});
|
| 838 |
+
|
| 839 |
+
// Sort by distance
|
| 840 |
+
let sortedPoints = [...distances].sort(
|
| 841 |
+
(a, b) => a.distance - b.distance
|
| 842 |
+
);
|
| 843 |
+
|
| 844 |
+
// Draw data points
|
| 845 |
+
dataPoints.forEach((point) => {
|
| 846 |
+
drawPoint(ennCtx, point.x, point.y, "#3498db", point.label);
|
| 847 |
+
});
|
| 848 |
+
|
| 849 |
+
// Draw query point
|
| 850 |
+
drawPoint(
|
| 851 |
+
ennCtx,
|
| 852 |
+
queryPoint.x,
|
| 853 |
+
queryPoint.y,
|
| 854 |
+
"#e74c3c",
|
| 855 |
+
queryPoint.label,
|
| 856 |
+
12
|
| 857 |
+
);
|
| 858 |
+
|
| 859 |
+
// Step-specific rendering
|
| 860 |
+
if (step >= 1) {
|
| 861 |
+
// Draw distance lines to all points
|
| 862 |
+
distances.forEach((point) => {
|
| 863 |
+
drawLine(
|
| 864 |
+
ennCtx,
|
| 865 |
+
queryPoint.x,
|
| 866 |
+
queryPoint.y,
|
| 867 |
+
point.x,
|
| 868 |
+
point.y,
|
| 869 |
+
"#aaa",
|
| 870 |
+
[3, 3]
|
| 871 |
+
);
|
| 872 |
+
|
| 873 |
+
// Draw distance value
|
| 874 |
+
const midX = (queryPoint.x + point.x) / 2;
|
| 875 |
+
const midY = (queryPoint.y + point.y) / 2;
|
| 876 |
+
ennCtx.fillStyle = "#555";
|
| 877 |
+
ennCtx.font = "11px Arial";
|
| 878 |
+
ennCtx.textAlign = "center";
|
| 879 |
+
ennCtx.fillText(point.distance.toFixed(1), midX, midY);
|
| 880 |
+
});
|
| 881 |
+
}
|
| 882 |
+
|
| 883 |
+
if (step >= 2) {
|
| 884 |
+
// Visualize sorting by distance
|
| 885 |
+
let yPos = 20;
|
| 886 |
+
ennCtx.fillStyle = "#333";
|
| 887 |
+
ennCtx.font = "12px Arial";
|
| 888 |
+
ennCtx.textAlign = "left";
|
| 889 |
+
ennCtx.fillText("Sorted by distance:", 10, yPos);
|
| 890 |
+
|
| 891 |
+
for (let i = 0; i < Math.min(5, sortedPoints.length); i++) {
|
| 892 |
+
yPos += 15;
|
| 893 |
+
ennCtx.fillText(
|
| 894 |
+
`${i + 1}. ${sortedPoints[i].label} (${sortedPoints[
|
| 895 |
+
i
|
| 896 |
+
].distance.toFixed(1)})`,
|
| 897 |
+
15,
|
| 898 |
+
yPos
|
| 899 |
+
);
|
| 900 |
+
}
|
| 901 |
+
}
|
| 902 |
+
|
| 903 |
+
if (step >= 3) {
|
| 904 |
+
// Highlight nearest neighbor(s)
|
| 905 |
+
const nearest = sortedPoints[0];
|
| 906 |
+
drawPoint(
|
| 907 |
+
ennCtx,
|
| 908 |
+
nearest.x,
|
| 909 |
+
nearest.y,
|
| 910 |
+
"#3498db",
|
| 911 |
+
nearest.label,
|
| 912 |
+
10,
|
| 913 |
+
"#2ecc71",
|
| 914 |
+
3
|
| 915 |
+
);
|
| 916 |
+
drawLine(
|
| 917 |
+
ennCtx,
|
| 918 |
+
queryPoint.x,
|
| 919 |
+
queryPoint.y,
|
| 920 |
+
nearest.x,
|
| 921 |
+
nearest.y,
|
| 922 |
+
"#2ecc71",
|
| 923 |
+
[],
|
| 924 |
+
2
|
| 925 |
+
);
|
| 926 |
+
|
| 927 |
+
// Draw threshold for the nearest distance
|
| 928 |
+
if (distanceMetric === "euclidean") {
|
| 929 |
+
ennCtx.beginPath();
|
| 930 |
+
ennCtx.arc(
|
| 931 |
+
queryPoint.x,
|
| 932 |
+
queryPoint.y,
|
| 933 |
+
nearest.distance,
|
| 934 |
+
0,
|
| 935 |
+
Math.PI * 2
|
| 936 |
+
);
|
| 937 |
+
ennCtx.strokeStyle = "rgba(231, 76, 60, 0.4)";
|
| 938 |
+
ennCtx.stroke();
|
| 939 |
+
ennCtx.fillStyle = "rgba(231, 76, 60, 0.05)";
|
| 940 |
+
ennCtx.fill();
|
| 941 |
+
} else if (distanceMetric === "manhattan") {
|
| 942 |
+
// Draw diamond shape
|
| 943 |
+
ennCtx.beginPath();
|
| 944 |
+
ennCtx.moveTo(queryPoint.x, queryPoint.y - nearest.distance);
|
| 945 |
+
ennCtx.lineTo(queryPoint.x + nearest.distance, queryPoint.y);
|
| 946 |
+
ennCtx.lineTo(queryPoint.x, queryPoint.y + nearest.distance);
|
| 947 |
+
ennCtx.lineTo(queryPoint.x - nearest.distance, queryPoint.y);
|
| 948 |
+
ennCtx.closePath();
|
| 949 |
+
ennCtx.strokeStyle = "rgba(231, 76, 60, 0.4)";
|
| 950 |
+
ennCtx.stroke();
|
| 951 |
+
ennCtx.fillStyle = "rgba(231, 76, 60, 0.05)";
|
| 952 |
+
ennCtx.fill();
|
| 953 |
+
} else if (distanceMetric === "cosine") {
|
| 954 |
+
// Complicated to visualize in 2D space, show a text note
|
| 955 |
+
ennCtx.fillStyle = "rgba(231, 76, 60, 0.7)";
|
| 956 |
+
ennCtx.fillText(
|
| 957 |
+
"Cosine similarity measures angle between vectors",
|
| 958 |
+
250,
|
| 959 |
+
30
|
| 960 |
+
);
|
| 961 |
+
ennCtx.fillText("smaller angle = more similar", 250, 45);
|
| 962 |
+
}
|
| 963 |
+
}
|
| 964 |
+
|
| 965 |
+
// Update step description
|
| 966 |
+
updateENNStepInfo(step, distanceMetric);
|
| 967 |
+
}
|
| 968 |
+
|
| 969 |
+
// ANN Search visualization
|
| 970 |
+
function renderANNSearch() {
|
| 971 |
+
const algorithm = annAlgorithmSelect.value;
|
| 972 |
+
const step = parseInt(annStepSelect.value);
|
| 973 |
+
|
| 974 |
+
// Clear canvas
|
| 975 |
+
annCtx.clearRect(0, 0, annCanvas.width, annCanvas.height);
|
| 976 |
+
|
| 977 |
+
// Draw grid
|
| 978 |
+
drawGrid(annCtx);
|
| 979 |
+
|
| 980 |
+
// Draw data points
|
| 981 |
+
dataPoints.forEach((point) => {
|
| 982 |
+
drawPoint(annCtx, point.x, point.y, "#3498db", point.label);
|
| 983 |
+
});
|
| 984 |
+
|
| 985 |
+
// Draw query point
|
| 986 |
+
drawPoint(
|
| 987 |
+
annCtx,
|
| 988 |
+
queryPoint.x,
|
| 989 |
+
queryPoint.y,
|
| 990 |
+
"#e74c3c",
|
| 991 |
+
queryPoint.label,
|
| 992 |
+
12
|
| 993 |
+
);
|
| 994 |
+
|
| 995 |
+
// Step-specific rendering based on algorithm
|
| 996 |
+
if (algorithm === "hnsw") {
|
| 997 |
+
renderHNSW(annCtx, step);
|
| 998 |
+
} else if (algorithm === "pq") {
|
| 999 |
+
renderProductQuantization(annCtx, step);
|
| 1000 |
+
} else if (algorithm === "lsh") {
|
| 1001 |
+
renderLSH(annCtx, step);
|
| 1002 |
+
}
|
| 1003 |
+
|
| 1004 |
+
// Update step description
|
| 1005 |
+
updateANNStepInfo(step, algorithm);
|
| 1006 |
+
}
|
| 1007 |
+
|
| 1008 |
+
// Semantic Search visualization
|
| 1009 |
+
function renderSemanticSearch() {
|
| 1010 |
+
const model = semanticModelSelect.value;
|
| 1011 |
+
const step = parseInt(semanticStepSelect.value);
|
| 1012 |
+
|
| 1013 |
+
// Clear canvas
|
| 1014 |
+
semanticCtx.clearRect(
|
| 1015 |
+
0,
|
| 1016 |
+
0,
|
| 1017 |
+
semanticCanvas.width,
|
| 1018 |
+
semanticCanvas.height
|
| 1019 |
+
);
|
| 1020 |
+
|
| 1021 |
+
if (step === 0) {
|
| 1022 |
+
// Show text documents
|
| 1023 |
+
drawTextDocuments(semanticCtx, semanticDocs, semanticQuery);
|
| 1024 |
+
} else {
|
| 1025 |
+
// Draw embedding space
|
| 1026 |
+
drawGrid(semanticCtx);
|
| 1027 |
+
|
| 1028 |
+
// Draw document embeddings (2D projection)
|
| 1029 |
+
semanticDocs.forEach((doc) => {
|
| 1030 |
+
// Scale to canvas
|
| 1031 |
+
const x = doc.embedding[0] * 400 + 30;
|
| 1032 |
+
const y = (1 - doc.embedding[1]) * 250 + 20;
|
| 1033 |
+
drawPoint(semanticCtx, x, y, "#3498db", `D${doc.id}`);
|
| 1034 |
+
});
|
| 1035 |
+
|
| 1036 |
+
// Draw query embedding
|
| 1037 |
+
const qx = semanticQuery.embedding[0] * 400 + 30;
|
| 1038 |
+
const qy = (1 - semanticQuery.embedding[1]) * 250 + 20;
|
| 1039 |
+
drawPoint(semanticCtx, qx, qy, "#e74c3c", "Q", 12);
|
| 1040 |
+
|
| 1041 |
+
if (step >= 2) {
|
| 1042 |
+
// Calculate similarities
|
| 1043 |
+
const similarities = semanticDocs
|
| 1044 |
+
.map((doc) => ({
|
| 1045 |
+
...doc,
|
| 1046 |
+
similarity: cosineSimilarity(
|
| 1047 |
+
doc.embedding,
|
| 1048 |
+
semanticQuery.embedding
|
| 1049 |
+
),
|
| 1050 |
+
}))
|
| 1051 |
+
.sort((a, b) => b.similarity - a.similarity);
|
| 1052 |
+
|
| 1053 |
+
// Draw lines to most similar docs
|
| 1054 |
+
for (let i = 0; i < 3; i++) {
|
| 1055 |
+
const doc = similarities[i];
|
| 1056 |
+
const dx = doc.embedding[0] * 400 + 30;
|
| 1057 |
+
const dy = (1 - doc.embedding[1]) * 250 + 20;
|
| 1058 |
+
|
| 1059 |
+
const lineWidth = 3 - i;
|
| 1060 |
+
drawLine(semanticCtx, qx, qy, dx, dy, "#2ecc71", [], lineWidth);
|
| 1061 |
+
|
| 1062 |
+
// Highlight the similar document
|
| 1063 |
+
drawPoint(
|
| 1064 |
+
semanticCtx,
|
| 1065 |
+
dx,
|
| 1066 |
+
dy,
|
| 1067 |
+
"#3498db",
|
| 1068 |
+
`D${doc.id}`,
|
| 1069 |
+
10,
|
| 1070 |
+
"#2ecc71",
|
| 1071 |
+
2
|
| 1072 |
+
);
|
| 1073 |
+
|
| 1074 |
+
// Show similarity score
|
| 1075 |
+
const midX = (qx + dx) / 2;
|
| 1076 |
+
const midY = (qy + dy) / 2 - 10;
|
| 1077 |
+
semanticCtx.fillStyle = "#555";
|
| 1078 |
+
semanticCtx.font = "11px Arial";
|
| 1079 |
+
semanticCtx.textAlign = "center";
|
| 1080 |
+
semanticCtx.fillText(doc.similarity.toFixed(2), midX, midY);
|
| 1081 |
+
}
|
| 1082 |
+
|
| 1083 |
+
if (step >= 3) {
|
| 1084 |
+
// Display top results
|
| 1085 |
+
let yPos = 20;
|
| 1086 |
+
semanticCtx.fillStyle = "#333";
|
| 1087 |
+
semanticCtx.font = "12px Arial";
|
| 1088 |
+
semanticCtx.textAlign = "left";
|
| 1089 |
+
semanticCtx.fillText("Top matches:", 10, yPos);
|
| 1090 |
+
|
| 1091 |
+
for (let i = 0; i < Math.min(3, similarities.length); i++) {
|
| 1092 |
+
yPos += 15;
|
| 1093 |
+
semanticCtx.fillText(
|
| 1094 |
+
`${similarities[i].text} (${similarities[
|
| 1095 |
+
i
|
| 1096 |
+
].similarity.toFixed(2)})`,
|
| 1097 |
+
15,
|
| 1098 |
+
yPos
|
| 1099 |
+
);
|
| 1100 |
+
}
|
| 1101 |
+
}
|
| 1102 |
+
}
|
| 1103 |
+
}
|
| 1104 |
+
|
| 1105 |
+
// Update step description
|
| 1106 |
+
updateSemanticStepInfo(step, model);
|
| 1107 |
+
}
|
| 1108 |
+
|
| 1109 |
+
// Sparse Vector Search visualization
|
| 1110 |
+
function renderSparseSearch() {
|
| 1111 |
+
const model = sparseModelSelect.value;
|
| 1112 |
+
const step = parseInt(sparseStepSelect.value);
|
| 1113 |
+
|
| 1114 |
+
// Clear canvas
|
| 1115 |
+
sparseCtx.clearRect(0, 0, sparseCanvas.width, sparseCanvas.height);
|
| 1116 |
+
|
| 1117 |
+
if (step === 0) {
|
| 1118 |
+
// Show text documents with highlighted tokens
|
| 1119 |
+
drawTokenizedDocuments(sparseCtx, sparseVectors, sparseQuery);
|
| 1120 |
+
} else {
|
| 1121 |
+
// Draw sparse vectors visualization
|
| 1122 |
+
drawSparseVectors(sparseCtx, sparseVectors, sparseQuery, step, model);
|
| 1123 |
+
|
| 1124 |
+
if (step >= 2) {
|
| 1125 |
+
// Calculate matching scores
|
| 1126 |
+
const matches = sparseVectors
|
| 1127 |
+
.map((doc) => {
|
| 1128 |
+
let score = 0;
|
| 1129 |
+
for (let i = 0; i < doc.vector.length; i++) {
|
| 1130 |
+
score += doc.vector[i] * sparseQuery.vector[i];
|
| 1131 |
+
}
|
| 1132 |
+
return { ...doc, score };
|
| 1133 |
+
})
|
| 1134 |
+
.sort((a, b) => b.score - a.score);
|
| 1135 |
+
|
| 1136 |
+
if (step >= 3) {
|
| 1137 |
+
// Display top results
|
| 1138 |
+
let yPos = 20;
|
| 1139 |
+
sparseCtx.fillStyle = "#333";
|
| 1140 |
+
sparseCtx.font = "12px Arial";
|
| 1141 |
+
sparseCtx.textAlign = "left";
|
| 1142 |
+
sparseCtx.fillText("Top matches:", 300, yPos);
|
| 1143 |
+
|
| 1144 |
+
for (let i = 0; i < Math.min(3, matches.length); i++) {
|
| 1145 |
+
yPos += 15;
|
| 1146 |
+
sparseCtx.fillText(
|
| 1147 |
+
`${matches[i].text} (${matches[i].score.toFixed(2)})`,
|
| 1148 |
+
300,
|
| 1149 |
+
yPos
|
| 1150 |
+
);
|
| 1151 |
+
}
|
| 1152 |
+
}
|
| 1153 |
+
}
|
| 1154 |
+
}
|
| 1155 |
+
|
| 1156 |
+
// Update step description
|
| 1157 |
+
updateSparseStepInfo(step, model);
|
| 1158 |
+
}
|
| 1159 |
+
|
| 1160 |
+
// Algorithm-specific renderers for ANN
|
| 1161 |
+
function renderHNSW(ctx, step) {
|
| 1162 |
+
if (step >= 1) {
|
| 1163 |
+
// Draw HNSW layers
|
| 1164 |
+
ctx.strokeStyle = "#f39c12";
|
| 1165 |
+
ctx.lineWidth = 1;
|
| 1166 |
+
|
| 1167 |
+
// Top layer (sparse connections)
|
| 1168 |
+
const topLayer = [dataPoints[2], dataPoints[4], dataPoints[7]];
|
| 1169 |
+
topLayer.forEach((p1, i) => {
|
| 1170 |
+
topLayer.forEach((p2, j) => {
|
| 1171 |
+
if (i !== j) {
|
| 1172 |
+
drawLine(ctx, p1.x, p1.y, p2.x, p2.y, "#f39c12", [2, 2], 1);
|
| 1173 |
+
}
|
| 1174 |
+
});
|
| 1175 |
+
});
|
| 1176 |
+
|
| 1177 |
+
// Middle layer (more connections)
|
| 1178 |
+
if (step >= 2) {
|
| 1179 |
+
const midLayer = [
|
| 1180 |
+
dataPoints[1],
|
| 1181 |
+
dataPoints[2],
|
| 1182 |
+
dataPoints[4],
|
| 1183 |
+
dataPoints[6],
|
| 1184 |
+
dataPoints[7],
|
| 1185 |
+
];
|
| 1186 |
+
midLayer.forEach((p1, i) => {
|
| 1187 |
+
let connections = 0;
|
| 1188 |
+
midLayer.forEach((p2, j) => {
|
| 1189 |
+
if (i !== j && connections < 3) {
|
| 1190 |
+
drawLine(ctx, p1.x, p1.y, p2.x, p2.y, "#f39c12", [], 1);
|
| 1191 |
+
connections++;
|
| 1192 |
+
}
|
| 1193 |
+
});
|
| 1194 |
+
});
|
| 1195 |
+
|
| 1196 |
+
// Entry point search
|
| 1197 |
+
const entryPoint = dataPoints[4]; // An arbitrary entry point - Error is solved
|
| 1198 |
+
drawPoint(
|
| 1199 |
+
ctx,
|
| 1200 |
+
entryPoint.x,
|
| 1201 |
+
entryPoint.y,
|
| 1202 |
+
"#3498db",
|
| 1203 |
+
entryPoint.label,
|
| 1204 |
+
10,
|
| 1205 |
+
"#f39c12",
|
| 1206 |
+
2
|
| 1207 |
+
);
|
| 1208 |
+
drawLine(
|
| 1209 |
+
ctx,
|
| 1210 |
+
queryPoint.x,
|
| 1211 |
+
queryPoint.y,
|
| 1212 |
+
entryPoint.x,
|
| 1213 |
+
entryPoint.y,
|
| 1214 |
+
"#f39c12",
|
| 1215 |
+
[],
|
| 1216 |
+
2
|
| 1217 |
+
);
|
| 1218 |
+
}
|
| 1219 |
+
|
| 1220 |
+
if (step >= 3) {
|
| 1221 |
+
// Show local greedy search path
|
| 1222 |
+
const searchPath = [
|
| 1223 |
+
dataPoints[4],
|
| 1224 |
+
dataPoints[7],
|
| 1225 |
+
dataPoints[6],
|
| 1226 |
+
dataPoints[2],
|
| 1227 |
+
];
|
| 1228 |
+
|
| 1229 |
+
for (let i = 0; i < searchPath.length - 1; i++) {
|
| 1230 |
+
const p1 = searchPath[i];
|
| 1231 |
+
const p2 = searchPath[i + 1];
|
| 1232 |
+
drawLine(ctx, p1.x, p1.y, p2.x, p2.y, "#e74c3c", [], 2);
|
| 1233 |
+
|
| 1234 |
+
if (i < searchPath.length - 2) {
|
| 1235 |
+
drawPoint(
|
| 1236 |
+
ctx,
|
| 1237 |
+
p1.x,
|
| 1238 |
+
p1.y,
|
| 1239 |
+
"#3498db",
|
| 1240 |
+
p1.label,
|
| 1241 |
+
10,
|
| 1242 |
+
"#f39c3c",
|
| 1243 |
+
2
|
| 1244 |
+
);
|
| 1245 |
+
}
|
| 1246 |
+
}
|
| 1247 |
+
|
| 1248 |
+
// Final result
|
| 1249 |
+
const nearest = dataPoints[2];
|
| 1250 |
+
drawPoint(
|
| 1251 |
+
ctx,
|
| 1252 |
+
nearest.x,
|
| 1253 |
+
nearest.y,
|
| 1254 |
+
"#3498db",
|
| 1255 |
+
nearest.label,
|
| 1256 |
+
10,
|
| 1257 |
+
"#2ecc71",
|
| 1258 |
+
3
|
| 1259 |
+
);
|
| 1260 |
+
drawLine(
|
| 1261 |
+
ctx,
|
| 1262 |
+
queryPoint.x,
|
| 1263 |
+
queryPoint.y,
|
| 1264 |
+
nearest.x,
|
| 1265 |
+
nearest.y,
|
| 1266 |
+
"#2ecc71",
|
| 1267 |
+
[],
|
| 1268 |
+
2
|
| 1269 |
+
);
|
| 1270 |
+
}
|
| 1271 |
+
}
|
| 1272 |
+
}
|
| 1273 |
+
|
| 1274 |
+
function renderProductQuantization(ctx, step) {
|
| 1275 |
+
if (step >= 1) {
|
| 1276 |
+
// Draw PQ centroids and quantized regions
|
| 1277 |
+
|
| 1278 |
+
// Split canvas into 4 regions (simple quantization visualization)
|
| 1279 |
+
ctx.strokeStyle = "#f39c12";
|
| 1280 |
+
ctx.lineWidth = 2;
|
| 1281 |
+
ctx.setLineDash([]);
|
| 1282 |
+
|
| 1283 |
+
// Vertical split
|
| 1284 |
+
ctx.beginPath();
|
| 1285 |
+
ctx.moveTo(ennCanvas.width / 2, 0);
|
| 1286 |
+
ctx.lineTo(ennCanvas.width / 2, ennCanvas.height);
|
| 1287 |
+
ctx.stroke();
|
| 1288 |
+
|
| 1289 |
+
// Horizontal split
|
| 1290 |
+
ctx.beginPath();
|
| 1291 |
+
ctx.moveTo(0, ennCanvas.height / 2);
|
| 1292 |
+
ctx.lineTo(ennCanvas.width, ennCanvas.height / 2);
|
| 1293 |
+
ctx.stroke();
|
| 1294 |
+
|
| 1295 |
+
// Label regions
|
| 1296 |
+
ctx.fillStyle = "#f39c12";
|
| 1297 |
+
ctx.font = "12px Arial";
|
| 1298 |
+
ctx.textAlign = "center";
|
| 1299 |
+
ctx.fillText("Region 1", ennCanvas.width / 4, ennCanvas.height / 4);
|
| 1300 |
+
ctx.fillText(
|
| 1301 |
+
"Region 2",
|
| 1302 |
+
(3 * ennCanvas.width) / 4,
|
| 1303 |
+
ennCanvas.height / 4
|
| 1304 |
+
);
|
| 1305 |
+
ctx.fillText(
|
| 1306 |
+
"Region 3",
|
| 1307 |
+
ennCanvas.width / 4,
|
| 1308 |
+
(3 * ennCanvas.height) / 4
|
| 1309 |
+
);
|
| 1310 |
+
ctx.fillText(
|
| 1311 |
+
"Region 4",
|
| 1312 |
+
(3 * ennCanvas.width) / 4,
|
| 1313 |
+
(3 * ennCanvas.height) / 4
|
| 1314 |
+
);
|
| 1315 |
+
|
| 1316 |
+
if (step >= 2) {
|
| 1317 |
+
// Identify query region
|
| 1318 |
+
let queryRegion;
|
| 1319 |
+
if (queryPoint.x < ennCanvas.width / 2) {
|
| 1320 |
+
if (queryPoint.y < ennCanvas.height / 2) {
|
| 1321 |
+
queryRegion = 1;
|
| 1322 |
+
} else {
|
| 1323 |
+
queryRegion = 3;
|
| 1324 |
+
}
|
| 1325 |
+
} else {
|
| 1326 |
+
if (queryPoint.y < ennCanvas.height / 2) {
|
| 1327 |
+
queryRegion = 2;
|
| 1328 |
+
} else {
|
| 1329 |
+
queryRegion = 4;
|
| 1330 |
+
}
|
| 1331 |
+
}
|
| 1332 |
+
|
| 1333 |
+
// Highlight query region
|
| 1334 |
+
ctx.fillStyle = "rgba(243, 156, 18, 0.1)";
|
| 1335 |
+
if (queryRegion === 1) {
|
| 1336 |
+
ctx.fillRect(0, 0, ennCanvas.width / 2, ennCanvas.height / 2);
|
| 1337 |
+
} else if (queryRegion === 2) {
|
| 1338 |
+
ctx.fillRect(
|
| 1339 |
+
ennCanvas.width / 2,
|
| 1340 |
+
0,
|
| 1341 |
+
ennCanvas.width / 2,
|
| 1342 |
+
ennCanvas.height / 2
|
| 1343 |
+
);
|
| 1344 |
+
} else if (queryRegion === 3) {
|
| 1345 |
+
ctx.fillRect(
|
| 1346 |
+
0,
|
| 1347 |
+
ennCanvas.height / 2,
|
| 1348 |
+
ennCanvas.width / 2,
|
| 1349 |
+
ennCanvas.height / 2
|
| 1350 |
+
);
|
| 1351 |
+
} else {
|
| 1352 |
+
ctx.fillRect(
|
| 1353 |
+
ennCanvas.width / 2,
|
| 1354 |
+
ennCanvas.height / 2,
|
| 1355 |
+
ennCanvas.width / 2,
|
| 1356 |
+
ennCanvas.height / 2
|
| 1357 |
+
);
|
| 1358 |
+
}
|
| 1359 |
+
|
| 1360 |
+
// Only search points in that region
|
| 1361 |
+
const pointsInRegion = dataPoints.filter((p) => {
|
| 1362 |
+
const region =
|
| 1363 |
+
p.x < ennCanvas.width / 2
|
| 1364 |
+
? p.y < ennCanvas.height / 2
|
| 1365 |
+
? 1
|
| 1366 |
+
: 3
|
| 1367 |
+
: p.y < ennCanvas.height / 2
|
| 1368 |
+
? 2
|
| 1369 |
+
: 4;
|
| 1370 |
+
return region === queryRegion;
|
| 1371 |
+
});
|
| 1372 |
+
|
| 1373 |
+
// Draw lines to only those points
|
| 1374 |
+
pointsInRegion.forEach((point) => {
|
| 1375 |
+
drawLine(
|
| 1376 |
+
ctx,
|
| 1377 |
+
queryPoint.x,
|
| 1378 |
+
queryPoint.y,
|
| 1379 |
+
point.x,
|
| 1380 |
+
point.y,
|
| 1381 |
+
"#aaa",
|
| 1382 |
+
[3, 3]
|
| 1383 |
+
);
|
| 1384 |
+
});
|
| 1385 |
+
}
|
| 1386 |
+
|
| 1387 |
+
if (step >= 3) {
|
| 1388 |
+
// Find approximated nearest (would be from the shortlisted region)
|
| 1389 |
+
const distances = dataPoints.map((point) => ({
|
| 1390 |
+
...point,
|
| 1391 |
+
distance: euclideanDistance(point, queryPoint),
|
| 1392 |
+
}));
|
| 1393 |
+
|
| 1394 |
+
// Filter to correct region first
|
| 1395 |
+
let queryRegion;
|
| 1396 |
+
if (queryPoint.x < ennCanvas.width / 2) {
|
| 1397 |
+
if (queryPoint.y < ennCanvas.height / 2) {
|
| 1398 |
+
queryRegion = 1;
|
| 1399 |
+
} else {
|
| 1400 |
+
queryRegion = 3;
|
| 1401 |
+
}
|
| 1402 |
+
} else {
|
| 1403 |
+
if (queryPoint.y < ennCanvas.height / 2) {
|
| 1404 |
+
queryRegion = 2;
|
| 1405 |
+
} else {
|
| 1406 |
+
queryRegion = 4;
|
| 1407 |
+
}
|
| 1408 |
+
}
|
| 1409 |
+
|
| 1410 |
+
const pointsInRegion = distances.filter((p) => {
|
| 1411 |
+
const region =
|
| 1412 |
+
p.x < ennCanvas.width / 2
|
| 1413 |
+
? p.y < ennCanvas.height / 2
|
| 1414 |
+
? 1
|
| 1415 |
+
: 3
|
| 1416 |
+
: p.y < ennCanvas.height / 2
|
| 1417 |
+
? 2
|
| 1418 |
+
: 4;
|
| 1419 |
+
return region === queryRegion;
|
| 1420 |
+
});
|
| 1421 |
+
|
| 1422 |
+
// Sort to find nearest in region
|
| 1423 |
+
const nearest = pointsInRegion.sort(
|
| 1424 |
+
(a, b) => a.distance - b.distance
|
| 1425 |
+
)[0];
|
| 1426 |
+
|
| 1427 |
+
// Highlight approximate nearest neighbor
|
| 1428 |
+
drawPoint(
|
| 1429 |
+
ctx,
|
| 1430 |
+
nearest.x,
|
| 1431 |
+
nearest.y,
|
| 1432 |
+
"#3498db",
|
| 1433 |
+
nearest.label,
|
| 1434 |
+
10,
|
| 1435 |
+
"#2ecc71",
|
| 1436 |
+
3
|
| 1437 |
+
);
|
| 1438 |
+
drawLine(
|
| 1439 |
+
ctx,
|
| 1440 |
+
queryPoint.x,
|
| 1441 |
+
queryPoint.y,
|
| 1442 |
+
nearest.x,
|
| 1443 |
+
nearest.y,
|
| 1444 |
+
"#2ecc71",
|
| 1445 |
+
[],
|
| 1446 |
+
2
|
| 1447 |
+
);
|
| 1448 |
+
|
| 1449 |
+
// Check if it's actually the true nearest neighbor
|
| 1450 |
+
const trueNearest = distances.sort(
|
| 1451 |
+
(a, b) => a.distance - b.distance
|
| 1452 |
+
)[0];
|
| 1453 |
+
if (nearest.id !== trueNearest.id) {
|
| 1454 |
+
// Show actual nearest as reference
|
| 1455 |
+
drawPoint(
|
| 1456 |
+
ctx,
|
| 1457 |
+
trueNearest.x,
|
| 1458 |
+
trueNearest.y,
|
| 1459 |
+
"#3498db",
|
| 1460 |
+
trueNearest.label,
|
| 1461 |
+
10,
|
| 1462 |
+
"#e74c3c",
|
| 1463 |
+
2
|
| 1464 |
+
);
|
| 1465 |
+
drawLine(
|
| 1466 |
+
ctx,
|
| 1467 |
+
queryPoint.x,
|
| 1468 |
+
queryPoint.y,
|
| 1469 |
+
trueNearest.x,
|
| 1470 |
+
trueNearest.y,
|
| 1471 |
+
"#e74c3c",
|
| 1472 |
+
[5, 5],
|
| 1473 |
+
1
|
| 1474 |
+
);
|
| 1475 |
+
|
| 1476 |
+
ctx.fillStyle = "#e74c3c";
|
| 1477 |
+
ctx.font = "12px Arial";
|
| 1478 |
+
ctx.textAlign = "left";
|
| 1479 |
+
ctx.fillText("Approximation error", 10, 20);
|
| 1480 |
+
ctx.fillText(`True nearest: ${trueNearest.label}`, 10, 35);
|
| 1481 |
+
} else {
|
| 1482 |
+
ctx.fillStyle = "#2ecc71";
|
| 1483 |
+
ctx.font = "12px Arial";
|
| 1484 |
+
ctx.textAlign = "left";
|
| 1485 |
+
ctx.fillText("Correct match", 10, 20);
|
| 1486 |
+
}
|
| 1487 |
+
}
|
| 1488 |
+
}
|
| 1489 |
+
}
|
| 1490 |
+
|
| 1491 |
+
// Helper functions for visualizations
|
| 1492 |
+
function drawGrid(ctx) {
|
| 1493 |
+
ctx.strokeStyle = "#e0e0e0";
|
| 1494 |
+
ctx.lineWidth = 0.5;
|
| 1495 |
+
|
| 1496 |
+
// Vertical lines
|
| 1497 |
+
for (let x = 0; x < ctx.canvas.width; x += 40) {
|
| 1498 |
+
ctx.beginPath();
|
| 1499 |
+
ctx.moveTo(x, 0);
|
| 1500 |
+
ctx.lineTo(x, ctx.canvas.height);
|
| 1501 |
+
ctx.stroke();
|
| 1502 |
+
}
|
| 1503 |
+
|
| 1504 |
+
// Horizontal lines
|
| 1505 |
+
for (let y = 0; y < ctx.canvas.height; y += 40) {
|
| 1506 |
+
ctx.beginPath();
|
| 1507 |
+
ctx.moveTo(0, y);
|
| 1508 |
+
ctx.lineTo(ctx.canvas.width, y);
|
| 1509 |
+
ctx.stroke();
|
| 1510 |
+
}
|
| 1511 |
+
}
|
| 1512 |
+
|
| 1513 |
+
function drawPoint(
|
| 1514 |
+
ctx,
|
| 1515 |
+
x,
|
| 1516 |
+
y,
|
| 1517 |
+
color,
|
| 1518 |
+
label,
|
| 1519 |
+
radius = 8,
|
| 1520 |
+
strokeColor = "#333",
|
| 1521 |
+
strokeWidth = 1
|
| 1522 |
+
) {
|
| 1523 |
+
ctx.beginPath();
|
| 1524 |
+
ctx.arc(x, y, radius, 0, Math.PI * 2);
|
| 1525 |
+
ctx.fillStyle = color;
|
| 1526 |
+
ctx.fill();
|
| 1527 |
+
ctx.strokeStyle = strokeColor;
|
| 1528 |
+
ctx.lineWidth = strokeWidth;
|
| 1529 |
+
ctx.stroke();
|
| 1530 |
+
|
| 1531 |
+
// Label
|
| 1532 |
+
ctx.fillStyle = "#333";
|
| 1533 |
+
ctx.font = "12px Arial";
|
| 1534 |
+
ctx.textAlign = "center";
|
| 1535 |
+
ctx.fillText(label, x, y - radius - 5);
|
| 1536 |
+
}
|
| 1537 |
+
|
| 1538 |
+
function drawLine(
|
| 1539 |
+
ctx,
|
| 1540 |
+
x1,
|
| 1541 |
+
y1,
|
| 1542 |
+
x2,
|
| 1543 |
+
y2,
|
| 1544 |
+
color = "#333",
|
| 1545 |
+
dash = [],
|
| 1546 |
+
width = 1
|
| 1547 |
+
) {
|
| 1548 |
+
ctx.beginPath();
|
| 1549 |
+
ctx.setLineDash(dash);
|
| 1550 |
+
ctx.strokeStyle = color;
|
| 1551 |
+
ctx.lineWidth = width;
|
| 1552 |
+
ctx.moveTo(x1, y1);
|
| 1553 |
+
ctx.lineTo(x2, y2);
|
| 1554 |
+
ctx.stroke();
|
| 1555 |
+
ctx.setLineDash([]);
|
| 1556 |
+
}
|
| 1557 |
+
|
| 1558 |
+
function drawTextDocuments(ctx, docs, query) {
|
| 1559 |
+
ctx.fillStyle = "#333";
|
| 1560 |
+
ctx.font = "14px Arial";
|
| 1561 |
+
ctx.textAlign = "left";
|
| 1562 |
+
|
| 1563 |
+
// Draw title
|
| 1564 |
+
ctx.fillText("Original Text Documents:", 20, 30);
|
| 1565 |
+
|
| 1566 |
+
// Draw documents
|
| 1567 |
+
let y = 60;
|
| 1568 |
+
docs.slice(0, 5).forEach((doc) => {
|
| 1569 |
+
ctx.fillStyle = "#3498db";
|
| 1570 |
+
ctx.fillText(`D${doc.id}: ${doc.text}`, 20, y);
|
| 1571 |
+
y += 25;
|
| 1572 |
+
});
|
| 1573 |
+
|
| 1574 |
+
// Draw query
|
| 1575 |
+
y += 20;
|
| 1576 |
+
ctx.fillStyle = "#e74c3c";
|
| 1577 |
+
ctx.fillText(`Query: "${query.text}"`, 20, y);
|
| 1578 |
+
|
| 1579 |
+
// Instructions
|
| 1580 |
+
y += 40;
|
| 1581 |
+
ctx.fillStyle = "#333";
|
| 1582 |
+
ctx.fillText(
|
| 1583 |
+
"Step 1: These documents will be converted to vector embeddings",
|
| 1584 |
+
20,
|
| 1585 |
+
y
|
| 1586 |
+
);
|
| 1587 |
+
ctx.fillText("that capture their semantic meaning.", 20, y + 20);
|
| 1588 |
+
}
|
| 1589 |
+
|
| 1590 |
+
function drawTokenizedDocuments(ctx, docs, query) {
|
| 1591 |
+
ctx.fillStyle = "#333";
|
| 1592 |
+
ctx.font = "14px Arial";
|
| 1593 |
+
ctx.textAlign = "left";
|
| 1594 |
+
|
| 1595 |
+
// Draw title
|
| 1596 |
+
ctx.fillText("Tokenized Documents:", 20, 30);
|
| 1597 |
+
|
| 1598 |
+
// Draw vocabulary
|
| 1599 |
+
ctx.fillText(
|
| 1600 |
+
"Vocabulary: dog, cat, train, pet, health, food, guide, home, behavior, puppy",
|
| 1601 |
+
20,
|
| 1602 |
+
50
|
| 1603 |
+
);
|
| 1604 |
+
|
| 1605 |
+
// Draw documents with highlighted tokens
|
| 1606 |
+
let y = 80;
|
| 1607 |
+
docs.slice(0, 5).forEach((doc) => {
|
| 1608 |
+
ctx.fillStyle = "#3498db";
|
| 1609 |
+
ctx.fillText(`D${doc.id}: ${doc.text}`, 20, y);
|
| 1610 |
+
|
| 1611 |
+
// Show token highlighting
|
| 1612 |
+
for (let i = 0; i < vocabulary.length; i++) {
|
| 1613 |
+
if (
|
| 1614 |
+
doc.vector[i] > 0 &&
|
| 1615 |
+
doc.text.toLowerCase().includes(vocabulary[i])
|
| 1616 |
+
) {
|
| 1617 |
+
const startX = 20 + ctx.measureText(`D${doc.id}: `).width;
|
| 1618 |
+
const wordStart = doc.text.toLowerCase().indexOf(vocabulary[i]);
|
| 1619 |
+
const prefix = doc.text.substring(0, wordStart);
|
| 1620 |
+
const prefixWidth = ctx.measureText(prefix).width;
|
| 1621 |
+
const wordWidth = ctx.measureText(vocabulary[i]).width;
|
| 1622 |
+
|
| 1623 |
+
ctx.fillStyle = "rgba(46, 204, 113, 0.3)";
|
| 1624 |
+
ctx.fillRect(startX + prefixWidth, y - 12, wordWidth, 15);
|
| 1625 |
+
}
|
| 1626 |
+
}
|
| 1627 |
+
|
| 1628 |
+
y += 25;
|
| 1629 |
+
});
|
| 1630 |
+
|
| 1631 |
+
// Draw query with highlighted tokens
|
| 1632 |
+
y += 20;
|
| 1633 |
+
ctx.fillStyle = "#e74c3c";
|
| 1634 |
+
ctx.fillText(`Query: "${query.text}"`, 20, y);
|
| 1635 |
+
|
| 1636 |
+
// Highlight query tokens
|
| 1637 |
+
for (let i = 0; i < vocabulary.length; i++) {
|
| 1638 |
+
if (
|
| 1639 |
+
query.vector[i] > 0 &&
|
| 1640 |
+
query.text.toLowerCase().includes(vocabulary[i])
|
| 1641 |
+
) {
|
| 1642 |
+
const startX = 20 + ctx.measureText(`Query: "`).width;
|
| 1643 |
+
const wordStart = query.text.toLowerCase().indexOf(vocabulary[i]);
|
| 1644 |
+
const prefix = query.text.substring(0, wordStart);
|
| 1645 |
+
const prefixWidth = ctx.measureText(prefix).width;
|
| 1646 |
+
const wordWidth = ctx.measureText(vocabulary[i]).width;
|
| 1647 |
+
|
| 1648 |
+
ctx.fillStyle = "rgba(231, 76, 60, 0.3)";
|
| 1649 |
+
ctx.fillRect(startX + prefixWidth, y - 12, wordWidth, 15);
|
| 1650 |
+
}
|
| 1651 |
+
}
|
| 1652 |
+
}
|
| 1653 |
+
|
| 1654 |
+
function drawSparseVectors(ctx, docs, query, step, model) {
|
| 1655 |
+
const barWidth = 15;
|
| 1656 |
+
const barSpacing = 5;
|
| 1657 |
+
const startX = 40;
|
| 1658 |
+
const startY = 220;
|
| 1659 |
+
const maxBarHeight = 100;
|
| 1660 |
+
|
| 1661 |
+
if (step >= 1) {
|
| 1662 |
+
// Draw vocabulary labels on x-axis
|
| 1663 |
+
ctx.fillStyle = "#333";
|
| 1664 |
+
ctx.font = "10px Arial";
|
| 1665 |
+
ctx.textAlign = "center";
|
| 1666 |
+
|
| 1667 |
+
vocabulary.forEach((word, i) => {
|
| 1668 |
+
const x = startX + i * (barWidth + barSpacing) + barWidth / 2;
|
| 1669 |
+
ctx.fillText(word, x, startY + 15);
|
| 1670 |
+
});
|
| 1671 |
+
|
| 1672 |
+
// Draw axis titles
|
| 1673 |
+
ctx.textAlign = "center";
|
| 1674 |
+
ctx.fillText("Vocabulary Terms", 230, startY + 30);
|
| 1675 |
+
|
| 1676 |
+
ctx.save();
|
| 1677 |
+
ctx.translate(15, 150);
|
| 1678 |
+
ctx.rotate(-Math.PI / 2);
|
| 1679 |
+
ctx.fillText("Term Weight", 0, 0);
|
| 1680 |
+
ctx.restore();
|
| 1681 |
+
|
| 1682 |
+
// Draw query vector
|
| 1683 |
+
ctx.fillStyle = "#333";
|
| 1684 |
+
ctx.font = "12px Arial";
|
| 1685 |
+
ctx.textAlign = "left";
|
| 1686 |
+
ctx.fillText("Query vector:", 20, 40);
|
| 1687 |
+
|
| 1688 |
+
query.vector.forEach((value, i) => {
|
| 1689 |
+
const x = startX + i * (barWidth + barSpacing);
|
| 1690 |
+
const barHeight = value * maxBarHeight;
|
| 1691 |
+
|
| 1692 |
+
ctx.fillStyle = value > 0 ? "#e74c3c" : "#f8f9fa";
|
| 1693 |
+
ctx.fillRect(x, startY - barHeight, barWidth, barHeight);
|
| 1694 |
+
|
| 1695 |
+
if (value > 0) {
|
| 1696 |
+
ctx.fillStyle = "#fff";
|
| 1697 |
+
ctx.textAlign = "center";
|
| 1698 |
+
ctx.font = "9px Arial";
|
| 1699 |
+
ctx.fillText(
|
| 1700 |
+
value.toFixed(1),
|
| 1701 |
+
x + barWidth / 2,
|
| 1702 |
+
startY - barHeight / 2
|
| 1703 |
+
);
|
| 1704 |
+
}
|
| 1705 |
+
|
| 1706 |
+
// Also draw mini version above
|
| 1707 |
+
const miniHeight = value * 20;
|
| 1708 |
+
ctx.fillStyle = value > 0 ? "#e74c3c" : "#f8f9fa";
|
| 1709 |
+
ctx.fillRect(x, 50, barWidth, miniHeight);
|
| 1710 |
+
});
|
| 1711 |
+
|
| 1712 |
+
if (step >= 2) {
|
| 1713 |
+
// Draw a document vector for comparison
|
| 1714 |
+
const matchingDoc = docs.find((d) => d.id === 1); // Dog training guide
|
| 1715 |
+
|
| 1716 |
+
ctx.fillStyle = "#333";
|
| 1717 |
+
ctx.font = "12px Arial";
|
| 1718 |
+
ctx.textAlign = "left";
|
| 1719 |
+
ctx.fillText(`Document: "${matchingDoc.text}"`, 20, 100);
|
| 1720 |
+
|
| 1721 |
+
matchingDoc.vector.forEach((value, i) => {
|
| 1722 |
+
const x = startX + i * (barWidth + barSpacing);
|
| 1723 |
+
const miniHeight = value * 20;
|
| 1724 |
+
|
| 1725 |
+
// Mini version above
|
| 1726 |
+
ctx.fillStyle = value > 0 ? "#3498db" : "#f8f9fa";
|
| 1727 |
+
ctx.fillRect(x, 110, barWidth, miniHeight);
|
| 1728 |
+
|
| 1729 |
+
// Highlight matching terms
|
| 1730 |
+
if (value > 0 && query.vector[i] > 0) {
|
| 1731 |
+
ctx.fillStyle = "#2ecc71";
|
| 1732 |
+
ctx.strokeStyle = "#2ecc71";
|
| 1733 |
+
ctx.lineWidth = 2;
|
| 1734 |
+
ctx.strokeRect(x, 50, barWidth, query.vector[i] * 20);
|
| 1735 |
+
ctx.strokeRect(x, 110, barWidth, miniHeight);
|
| 1736 |
+
|
| 1737 |
+
// Draw connection
|
| 1738 |
+
drawLine(
|
| 1739 |
+
ctx,
|
| 1740 |
+
x + barWidth / 2,
|
| 1741 |
+
50 + query.vector[i] * 20,
|
| 1742 |
+
x + barWidth / 2,
|
| 1743 |
+
110,
|
| 1744 |
+
"#2ecc71",
|
| 1745 |
+
[],
|
| 1746 |
+
1
|
| 1747 |
+
);
|
| 1748 |
+
}
|
| 1749 |
+
});
|
| 1750 |
+
|
| 1751 |
+
// Show dot product calculation
|
| 1752 |
+
let dotProduct = 0;
|
| 1753 |
+
for (let i = 0; i < query.vector.length; i++) {
|
| 1754 |
+
dotProduct += query.vector[i] * matchingDoc.vector[i];
|
| 1755 |
+
}
|
| 1756 |
+
|
| 1757 |
+
ctx.fillStyle = "#333";
|
| 1758 |
+
ctx.font = "12px Arial";
|
| 1759 |
+
ctx.textAlign = "left";
|
| 1760 |
+
ctx.fillText(`Matching score: ${dotProduct.toFixed(2)}`, 320, 100);
|
| 1761 |
+
}
|
| 1762 |
+
}
|
| 1763 |
+
}
|
| 1764 |
+
|
| 1765 |
+
// Update step descriptions
|
| 1766 |
+
function updateENNStepInfo(step, distanceMetric) {
|
| 1767 |
+
let title, description;
|
| 1768 |
+
|
| 1769 |
+
switch (step) {
|
| 1770 |
+
case 0:
|
| 1771 |
+
title = "Step 0: Data points";
|
| 1772 |
+
description =
|
| 1773 |
+
"Initial dataset with vectors in feature space. The query point (red) will be compared against all data points.";
|
| 1774 |
+
break;
|
| 1775 |
+
case 1:
|
| 1776 |
+
title = "Step 1: Calculate all distances";
|
| 1777 |
+
if (distanceMetric === "euclidean") {
|
| 1778 |
+
description =
|
| 1779 |
+
"Calculate Euclidean (L2) distance between query and every data point: d = √((x₂-x₁)² + (y₂-y₁)²).";
|
| 1780 |
+
} else if (distanceMetric === "manhattan") {
|
| 1781 |
+
description =
|
| 1782 |
+
"Calculate Manhattan (L1) distance between query and every data point: d = |x₂-x₁| + |y₂-y₁|.";
|
| 1783 |
+
} else {
|
| 1784 |
+
description =
|
| 1785 |
+
"Calculate Cosine similarity between query and data points: similarity = cos(θ) between vectors.";
|
| 1786 |
+
}
|
| 1787 |
+
break;
|
| 1788 |
+
case 2:
|
| 1789 |
+
title = "Step 2: Sort by distance";
|
| 1790 |
+
description =
|
| 1791 |
+
"Sort all data points by their distance to query point (ascending order for distance, descending for similarity).";
|
| 1792 |
+
break;
|
| 1793 |
+
case 3:
|
| 1794 |
+
title = "Step 3: Return nearest neighbors";
|
| 1795 |
+
description =
|
| 1796 |
+
"Return the k closest data points (here k=1). This approach guarantees finding the exact nearest neighbor.";
|
| 1797 |
+
break;
|
| 1798 |
+
}
|
| 1799 |
+
|
| 1800 |
+
ennStepTitle.textContent = title;
|
| 1801 |
+
ennStepDesc.textContent = description;
|
| 1802 |
+
}
|
| 1803 |
+
|
| 1804 |
+
function updateANNStepInfo(step, algorithm) {
|
| 1805 |
+
let title, description;
|
| 1806 |
+
|
| 1807 |
+
switch (step) {
|
| 1808 |
+
case 0:
|
| 1809 |
+
title = "Step 0: Indexed structure";
|
| 1810 |
+
if (algorithm === "hnsw") {
|
| 1811 |
+
description =
|
| 1812 |
+
"HNSW pre-organizes vectors into a navigable small world graph with multiple layers for efficient search.";
|
| 1813 |
+
} else if (algorithm === "pq") {
|
| 1814 |
+
description =
|
| 1815 |
+
"Product Quantization divides the vector space into smaller subspaces and quantizes each dimension group.";
|
| 1816 |
+
} else {
|
| 1817 |
+
description =
|
| 1818 |
+
"Locality-Sensitive Hashing uses hash functions that map similar vectors to the same buckets.";
|
| 1819 |
+
}
|
| 1820 |
+
break;
|
| 1821 |
+
case 1:
|
| 1822 |
+
title = "Step 1: Navigate to region";
|
| 1823 |
+
if (algorithm === "hnsw") {
|
| 1824 |
+
description =
|
| 1825 |
+
"Search begins at a random entry point in the top layer (sparse connections).";
|
| 1826 |
+
} else if (algorithm === "pq") {
|
| 1827 |
+
description =
|
| 1828 |
+
"The query is mapped to specific regions in each subspace based on quantized centroids.";
|
| 1829 |
+
} else {
|
| 1830 |
+
description =
|
| 1831 |
+
"Query vector is hashed to identify which bucket(s) to search.";
|
| 1832 |
+
}
|
| 1833 |
+
break;
|
| 1834 |
+
case 2:
|
| 1835 |
+
title = "Step 2: Local search";
|
| 1836 |
+
if (algorithm === "hnsw") {
|
| 1837 |
+
description =
|
| 1838 |
+
"Navigate through connections to find closer and closer neighbors, descending through layers.";
|
| 1839 |
+
} else if (algorithm === "pq") {
|
| 1840 |
+
description =
|
| 1841 |
+
"Compare only with points in the same or nearby quantized regions to limit search space.";
|
| 1842 |
+
} else {
|
| 1843 |
+
description =
|
| 1844 |
+
"Only compute distances for vectors in the same hash bucket, dramatically reducing comparisons.";
|
| 1845 |
+
}
|
| 1846 |
+
break;
|
| 1847 |
+
case 3:
|
| 1848 |
+
title = "Step 3: Return approximate NN";
|
| 1849 |
+
if (algorithm === "hnsw") {
|
| 1850 |
+
description =
|
| 1851 |
+
"Return the closest point found. May not be the true nearest neighbor, but usually very close.";
|
| 1852 |
+
} else if (algorithm === "pq") {
|
| 1853 |
+
description =
|
| 1854 |
+
"Approximates distances between query and dataset points. Fast but loses some precision.";
|
| 1855 |
+
} else {
|
| 1856 |
+
description =
|
| 1857 |
+
"If points fall into different buckets, LSH might miss true nearest neighbors (accuracy vs. speed tradeoff).";
|
| 1858 |
+
}
|
| 1859 |
+
break;
|
| 1860 |
+
}
|
| 1861 |
+
|
| 1862 |
+
annStepTitle.textContent = title;
|
| 1863 |
+
annStepDesc.textContent = description;
|
| 1864 |
+
}
|
| 1865 |
+
|
| 1866 |
+
function updateSemanticStepInfo(step, model) {
|
| 1867 |
+
let title, description;
|
| 1868 |
+
|
| 1869 |
+
switch (step) {
|
| 1870 |
+
case 0:
|
| 1871 |
+
title = "Step 0: Text documents";
|
| 1872 |
+
description = "Raw text data before encoding into vector space.";
|
| 1873 |
+
break;
|
| 1874 |
+
case 1:
|
| 1875 |
+
title = "Step 1: Generate embeddings";
|
| 1876 |
+
if (model === "bert") {
|
| 1877 |
+
description =
|
| 1878 |
+
"BERT creates dense vector embeddings (768 dimensions) that capture semantic meaning of text.";
|
| 1879 |
+
} else if (model === "use") {
|
| 1880 |
+
description =
|
| 1881 |
+
"Universal Sentence Encoder maps sentences to 512-dimensional vectors that capture meaning.";
|
| 1882 |
+
} else {
|
| 1883 |
+
description =
|
| 1884 |
+
"Domain-specific embeddings capture meaning relevant to particular fields or applications.";
|
| 1885 |
+
}
|
| 1886 |
+
break;
|
| 1887 |
+
case 2:
|
| 1888 |
+
title = "Step 2: Vector similarity search";
|
| 1889 |
+
description =
|
| 1890 |
+
"Calculate similarity (usually cosine) between query vector and document vectors.";
|
| 1891 |
+
break;
|
| 1892 |
+
case 3:
|
| 1893 |
+
title = "Step 3: Return relevant results";
|
| 1894 |
+
description =
|
| 1895 |
+
"Rank documents by similarity and return the most relevant. Results include semantic matches, not just exact keyword matches.";
|
| 1896 |
+
break;
|
| 1897 |
+
}
|
| 1898 |
+
|
| 1899 |
+
semanticStepTitle.textContent = title;
|
| 1900 |
+
semanticStepDesc.textContent = description;
|
| 1901 |
+
}
|
| 1902 |
+
|
| 1903 |
+
function updateSparseStepInfo(step, model) {
|
| 1904 |
+
let title, description;
|
| 1905 |
+
|
| 1906 |
+
switch (step) {
|
| 1907 |
+
case 0:
|
| 1908 |
+
title = "Step 0: Tokenized content";
|
| 1909 |
+
description =
|
| 1910 |
+
"Documents broken down into tokens (words/terms) before converting to sparse vector representation.";
|
| 1911 |
+
break;
|
| 1912 |
+
case 1:
|
| 1913 |
+
title = "Step 1: Create sparse vectors";
|
| 1914 |
+
if (model === "tfidf") {
|
| 1915 |
+
description =
|
| 1916 |
+
"TF-IDF weights tokens based on term frequency and inverse document frequency to emphasize distinctive terms.";
|
| 1917 |
+
} else if (model === "bm25") {
|
| 1918 |
+
description =
|
| 1919 |
+
"BM25 extends TF-IDF with better term saturation and document length normalization.";
|
| 1920 |
+
} else {
|
| 1921 |
+
description =
|
| 1922 |
+
"Hybrid representations combine sparse (keyword) and dense (semantic) vectors for better retrieval.";
|
| 1923 |
+
}
|
| 1924 |
+
break;
|
| 1925 |
+
case 2:
|
| 1926 |
+
title = "Step 2: Inverted index search";
|
| 1927 |
+
description =
|
| 1928 |
+
"Lookup only the specific terms present in the query, accessing posting lists through an inverted index.";
|
| 1929 |
+
break;
|
| 1930 |
+
case 3:
|
| 1931 |
+
title = "Step 3: Return matches";
|
| 1932 |
+
description =
|
| 1933 |
+
"Return documents with matching terms, ranked by relevance score. Very efficient for exact term matches.";
|
| 1934 |
+
break;
|
| 1935 |
+
}
|
| 1936 |
+
|
| 1937 |
+
sparseStepTitle.textContent = title;
|
| 1938 |
+
sparseStepDesc.textContent = description;
|
| 1939 |
+
}
|
| 1940 |
+
</script>
|
| 1941 |
+
</body>
|
| 1942 |
</html>
|