ARFF for natural language processing

I'm trying to take a set of reviews, and convert them into the ARFF format for use with WEKA. Unfortunately either I completely misunderstand how the for开发者_开发百科mat works, or I'll have to have an attribute for ALL possible words, then a presence indicator. Does anyone know a better way, or ideally have a sample ARFF file?

If you store the reviews in plain text files and different folders (positive and negative in your case) you can use TextDirectoryLoader.

You find this in the KnowledgeFlow application in Weka or from the command line. More info here: http://weka.wikispaces.com/ARFF+files+from+Text+Collections

Took a while to work out, but with this input.arff:

@relation text_files

@attribute review string
@attribute sentiment {0, 1}

@data
"this is some text", 1
"this is some more text", 1
"different stuff", 0

And this command:

java -classpath "C:\\Program Files\\Weka-3-6\\weka.jar" weka.filters.unsupervised.attribute.StringToWordVector -i input.arff -o output.arff

The following is produced:

@relation 'text_files-weka.filters.unsupervised.attribute.StringToWordVector-R1-W1000-prune-rate-1.0-N0-stemmerweka.core.stemmers.NullStemmer-M1-tokenizerweka.core.tokenizers.WordTokenizer -delimiters \" \\r\\n\\t.,;:\\\'\\\"()?!\"'

@attribute sentiment {0,1}
@attribute different numeric
@attribute is numeric
@attribute more numeric
@attribute some numeric
@attribute stuff numeric
@attribute text numeric
@attribute this numeric

@data

{0 1,2 1,4 1,6 1,7 1}
{0 1,2 1,3 1,4 1,6 1,7 1}
{1 1,5 1}