nat-initmat - initialize a sparse matrix with words co-occurrence.
nat-initmat <crp1> <crp2> [<exc1> <exc2>] <matrix>
This tool is used internally by
nat-these and is not intended to be used independently. Basically, this tool takes two corpora files created by
nat-pre and allocates a sparse matrix, where rows indexes correspond to word identifiers on the source corpus, and column indexes correspond to word identifiers on the target corpus. Cells count the words co-occurrence on the same sentence. The
matrix file is then created with the matrix information.
Optionally, you can pass to the system two exclude lists, as returned by the
nat-words2id tool. These words will be ignored, and counting will not be done for them.
The matrix is saved and can be processed later by EM-Algorithm methods IPFP (
nat-ipfp), Sample A (
nat-samplea) and Sample B (
nat-words2id, nat-pre, NATools documentation
Copyright (C)2002-2009 Alberto Simoes and Jose Joao Almeida Copyright (C)1998 Djoerd Hiemstra GNU GENERAL PUBLIC LICENSE (LGPL) Version 2 (June 1991)