I try to resolve use case which catch semantic emotional word on text.
I explan, I have a system of ticketing and I want to analyse the text of requester and the answer of technician for detecting a potential negative or positive sentiments.
But I don’t know where to start because with the type of analysis I have no key word weigth or I don’t know where find the data.
this workflow shows how to do sentiment analysis in KNIME.
In order to use it for your use-case, you will have to provide the texts along with a class (positive/negative sentiment) as training data.
I have seen the workflow but my document contains no positive or negative values.
I don’t understand how to classify my data on positive or negative sentiment.
First you load the reference dataset into KNIME. Also, you process your original text with a Bag Of Words node and a Term To String node. Then you join your text with the sentiment data, giving each word in your text a sentiment if available in the reference dataset. Finally you aggregate the sentiments with a Group By node (e.g. mean), so that you get a total sentiment of your document.
@arttom I updated my description: You just have one process: After you aggregated the sentiments for the document you are done. You have the sentiment (e.g. a number between -1 and +1) for the document. You don’t need any classification anymore.