LT-Suggest Helps Users to Find Even the Most Hidden Information

Our approximate indexing technology provides a very flexible and extremly fast fuzzy matching to large dictionaries, i.e. huge and synthetically structured suggest lists. It is based on a robust server/client architecture that makes deployment easy and scalable. At its core, a combination of n-gram and edit-distance algorithms is capable of covering all types of orthographical variation (misspellings, typos, different spelling conventions, garbled text input from OCR etc).

Typical application areas are:

  • Query recognition
  • Query repair / did-you-mean
  • Address matching
  • Address de-duplication
  • Text cleansing

As a type-ahead / autocomplete / suggest search interface, LT-Suggest comes in different flavors to exactly meet your users' needs

  • Fault-tolerant one-box suggest
  • Hierarchical / grouped suggest
  • Iterative suggest (combining several input fields in one suggest box)
  • Embedded suggest (synthetic generation of billions of possible suggest terms)
  • Suggest boxes as a replacement to search
  • Suggest boxes as an add-on to search

Features of LT-Suggest

  • Simple and straightforward to use (via command shell or Web-GUI)
  • Index is composed of key-value associations
  • Index preprocessing / normalization (handling of diacritics, spaces/dashes and other non-alphanumerical chars)
  • UTF-8 support for international charsets
  • Per-record aliases are supported
  • Per-record boosts are supported
  • Postfiltering of results
  • Geographical approximation is supported via a fast 2D-Index
  • Robust Server/Client architecture (HTTP requests) with optimized throughput (up to 20,000 req/s - depending on the index) via smart preforking mechanism