At LCI we developed the world first text classification engine based on a phono-semantic representation. This classifier has proven itself in many difficult cases and especially in a heavy-duty environment with large sets of examples. LCI's first product, a new generation layout image classifier and clustering engine, also won acclam from all our clients. It runs fast enough to keep pace with the hardware at a well known US professional scanner producer, sorting automatically the scanned documents in real time.

Our major achievement at LCI was C.A.R.E., the Content Addressable Record Extractor. C.A.R.E. uses fuzzy search in order to identify and match database records with noisy data originating via OCR from paper documents.

It works even for bad images (strong noise) and when the document's data barely match the database structure.

Fuzzy search is algorithmically much slower than exact string matching. Hence, LCI initiated the AMASS Consortium and received an FP6 EU Grant to build hardware support.

The dedicated FPGA platform performs the most time intensive part of the fuzzy search and extends the C.A.R.E. technology to other application fields: video search, semantic search, and genomic sequence search.

You can read more about the AMASS Project here:

• High performance search engine core

• Final Activity Report

• Extending the AMASS Platform

In 2009 R-EF GmbH delivered LCI (called then Kofax Development GmbH) the C.A.R.E. 2.0 library with full FPGA support, which is 50-100 times faster than the mainframe version.