"Sentiment Analysis similar to Vancouver Text Analytics Part 5 video"

bmartin

Hi,
I've been following the excellent Vancouver Blog Part 5 Automatic Classification of Documents Part 5 video.

I'm using RapdiMiner 5.1.003 and am attempting to do the following:

• Read from Oracle XE table containing 3 columns: ID, MESSAGE, and SENTIMENT_TYPE - to explain the ID is a unique identifier, the MESSAGE is a tweet and the SENTIMENT_TYPE describes whether the message is negative, positive, or neutral (i.e. a label)
• Convert from Nominal to Text
• Process documents from Data including transform case, stemming, tokenise
• Select Attribute and then Set role

What i would like to do is use the Set Role to set the ID and label (sentiment type) as is done in video. Then I could go and use X-Validation to train and predict the data etc and I would like to output the ID and sentiment_type back into the same database

However, when i run the program the Process Documents from Data outputs one column - seems to concatenate the ID, MESSAGE and sentiment type. The sentiment type (or ID) is not available for selection in the 'Set Role' operator.
Even when I click the 'add meta information' in the 'Process Documents from Data' operator it does the same.

Anyone know how to get around this issue?

Thanks in advance,
Brian

B_

This will help you get started. 

http://rapid-i.com/rapidforum/index.php/topic,2993.0.html

bmartin

Many thanks - very useful process to use...