The PruningRadixTrie is a revolutionary data structure that is derived from a radix trie, but is three orders of magnitude faster. It allows for pruning of the trie and early termination of the lookup, resulting in a significant reduction in lookup time for the top-k most relevant terms. This is particularly useful for auto-complete and prefix search in large dictionaries. The performance of the PruningRadixTrie is up to 1000x faster than an ordinary radix trie, making it ideal for serving thousands of users in parallel. The data structure is easy to use, with methods for adding terms, retrieving top-k terms, and reading/writing to disk. SeekStorm, the provider of this structure, offers high-performance search services and an API. Some noteworthy implementations of the PruningRadixTrie have been made in other programming languages, such as Python, Java, and Rust.
https://github.com/wolfgarbe/PruningRadixTrie