Evaluating Unstructured Text / Software Support Tickets

I am new to KNIME and Text Analytics in general so I don't know what can and cannot be done.

We'd like to use KNIME to evaluate software support tickets containing unstructured text. We would like to analyze approximately 3,000 tickets received per month to identify patterns and prioritize issues that are difficult to identify with manual analysis because customers don't use a common vocabulary to describe their issues.

We don't know if KNIME can do this. So far, I haven't found articles or videos descibing this use case.

I am hoping that someone in this forum is using KNIME for this purpose and can share their experience.

Thank you.

Scott C.



Hi Scott,

yes this is definitely what KNIME is designed to do. Do you already know about the text processing extension?

Best regards, Iris


Thank you for your reply. Yes. I know about the Text processing extension. I've created a few workflows using nodes from the Text Processing library to become more familiar with the KNIME workflow model. I'm encouraged but learning KNIME by watching videos and reading whitepapers without the ability to ask questions is going to take too long.

Assuming you are fluent in KNIME, may I ask how you learned it? Please don't tell me that you just picked it up all on your own without any help! :) 

I feel certain that many software companies are using KNIME to identify patterns in unstructured text from Support Tickets. These tickets contain a "gold mine" of information that we need to prioritize software enhancements and improvements.  I would think that every software company in the world, including KNIME, has this challenge / opportunity. None of the tutorials or videos on KNIME describe this application specifically. I wonder if it's just too difficult.

Over time, I may be able to stumble upon the right sequence of nodes for our application but I'm hoping to find a tutorial explaining which nodes are needed and why - hopefully with a screenshot of the workflow. I can hope can't I? With over 1,000 nodes, trial and error is going to take a long time!

We need a way to accurately, consistently and automatically determine what each support ticket is "about" - with the ability to create custom topics containing differently worded phrases that are actually describing the same topic.

We would also like to identify trends to show if our product improvements are being noticed by our customers.

It would be very helpful if KNIME was also able to flag tickets with high emotional content because certain types of problems are more frustrating to customers. The more emotional issues increase the risk of churn.

I'm just scratching the surface of our potential use of KNIME but I need a faster way to learn it.

Any suggestions?

Thank you.



I think I am not a good example.  I started as a student assistant working with the KNIME Analytics Plattform in... 2007(?)  At this point in time it was easy to know all the nodes. But right now I doubt there is someone who knows each node being availabe via the KNIME Analytics Plattform. Especially as there is quite some community developing is going on.

Our online training https://www.knime.org/knime-online-self-training was just recently published. This guides you pretty nice through our various materials.

On the other hand, we offer regular trainings in Zurich and at our KNIME Spring Summit there will be various more specific training courses.

Best regards, Iris

A post was split to a new topic: Text Analytics, Pre-processing, Visualization