"Text Mining from Database Column"

Hi,

I have looked for a solution to this and have yet to find it.  I have a large amount of text in a database table that I want to assign sentiment values.  I have labeled 2000+ separate text lines and want to use this to create the model (via SVM) and use this model to label the remaining unlabeled data.

So far, I have:

READ DATABASE => NORMALIZE TO TEXT (How do I specify the text field, so I use only that field in the Read Database query?)
=> DATA TO DOCUMENT => PROCESS DOCUMENTS => W-SMO => WRITE DATABASE

I see the one output is a model, where does this go?
Where do I pull in the model?
How can I properly label the labeled texts?  (Do I need a new table??)



Here's one attempt for building the model.

Read Database =>
Set Role =>
Set Role (2) =>
Nominal to Text =>
Replace Missing Values =>
W-SMO =>
Write Model => file.mod

I get the following error: The process will need more than the maximum amount of available memory...



Your insight is appreciated.

Thanks,
Bob