BST is a binary search tree. It is used for vocabulary. BST has no structural limitations, and therefore search / insert / delete is O (n) worst case.
The map [on stl] is also a dictionary and is actually a red-black tree [on stl]. this is a special kind of BST that has limitations on structures, which is why in the worst case, search / insert / delete is O (logn).
<x> hashing a hash table is another type of dictionary, the advantage of a hash table [with good hash functions] is O (1) average search / delete / insert time. however, the worst case is O (n), which happens if too many elements collide and / or when renaming is required [when the Load Balance is too high, we allocate a larger array and replay all the elements in is].
Tries are special for strings. all operators are O (S), where S is the length of the string. this is an advantage for others [when working with strings] - you still need to read the string, so the difficulty if a Map , for example, when working with strings, is actually O (S * n * logn).
when to use?
a Map [or any other balanced tree] should always be a better choice than a regular BST .
hash table is a good choice when you need a medium short time, but don't care that in some cases you will have a performance loss due to rephrasing, and in some cases there may be collisions.
Trie usually a good dictionary for strings.