what concept is used to develop dictionary

My understanding of the performance of general-purpose hash-tables (dictionaries): you can't get constant access time.
A simple counter example: If you are unfortunate enough that all keys map to the same "index" location, you have to do collision handling all entries.

If your hash table is contraint, e.g. a key word list for a computer language, you can tailor your hash algoritm (key calculation and hash size) so that you have a guaranted first hit or a failure.

I did that exercise once in the distant past in some Verilog compiler development (in C++). I had a set of key words, the key algoritm was such that the first 8 characters or so were mangled by simple operations into an integer.
With a perl script I calculated a hash table size once at development time by brute force such that each key modulo hash-table-size gave a unique index, and the table-size was in the accepted size range.

With that you had a first hit or none, and with the hit, the string had to be compared. Close enough to almost constant access time (worst case time: longest word to compare, where the last character differs to the longes keyword - neglectable).

Cheers

Andi