"SOLVED - Text Mining"

rarora

Dear experts,
I am new to RapidMiner and have gone through al the video's to come upto speed. I am trying to use RM to understand feedback given by our customers on our services. These feedback are  positve, negative and also areas for improvement. We want to classify these feedbacks into 6-7 classes.  Feedbacks are in individual text files - one file for one customer.

I am able to load the data, tokenize and do all preprocessing and generate n-grams (trigrams). It produces TF-IDF, it has a list of about 5 thousand n-grams (1-,2- and 3- grams). How do I tell RapidMinor what my classes are and which n-gram maps to that class.

Any suggestions/leads will be really helpful.

Thanks
Rajeev

MariusHelf

Hi,

you need labeled training data, that means that you have to go through as many feedbacks as possible manually and assign one of "positive", "negative" or "neutral" to them. If you have your data e.g. in an Excel file, you could add a column for that. Then, in RapidMiner you define that column as label and apply a normal learning schema, e.g. an SVM.
If your data is already labeled, you can of course skip the manual labeling process

Best,
Marius