FEX is a tool for extracting features from text. These features can be used to generate examples for use with Machine Learning software such as SNoW.
This page gives a brief outline of FEX. More technical details can be found in the user guide. Below you will find links which will allow you to download the code and view the software license. There are also links to detailed instructions on how to install and use FEX. You can also sign up for the FEX mailing lists and find annotated data sets for use in your experiments. We will soon add a tutorial that shows how FEX and SNoW can be used in some representative Machine Learning tasks.
FEX processes a body of data (corpus) using a set of commands which determine what features it can generate (the script). It builds (or updates) a lexicon, which maps features it generates from the text to integer numbers. FEX outputs examples (lists of integers corresponding to features in the lexicon), which will be labeled (also with integers) if the script contains the appropriate commands. These examples are deisgned to be compatible with SNoW.
FEX can process text in several different forms, including plain text, POS (Part-Of-Speech)-tagged text, and column format. These different formats represent text data at different stages of processing, with column format allowing a richer set of annotations than either plain or POS-tagged text. The details of each valid format can be found in the FEX user guide. Fex can easily be compiled on most UNIX systems. If you are working on a Win32 platform, we encourage you to use Cygwin to compile our software.
There are a variety of tools available on this site for converting raw text to
other appropriate formats, and there are some links to tagged corpora on the
links page.
Mailing Lists:
Participants:
Relevant Tools:
