You have been told correctly. Fortunately, the SAS Text Miner software does a lot of this work for you. We include text parsers that can identify language features in many different languages and also routines for efficiently transforming those features into variables that can be used in clustering and profiling to identify themes and in predictive models.
Your input data set should contain both the text fields and the scaled question responses. After the text mining node, you can add a clustering node and then a segment profile node. If you consider the clusters to contain or represent themes, then profiles plots and tree in the cluster node, and the more extensive reporting in the segment profile node will give you an idea how the text and question responses are related. You can do a lot with this type of analysis, so go have some fun.
-Dave