Text mining: Can responses be weighted differently?

Does anyone know if it is possible to weight specific responses differently? For example, my team had someone respond on behalf of a group of people where everyone else simply respondent from their own perspective. Hoping there is a way to weight and/or treat these responses differently in knime instead of potentially copy/pasting this response X times for the number of people it represents.

Thanks in advance.

Hi @Cyrprano

Could you give a bit more information?
How does the dat look like and your workflow?

My team has survey data in an excel file where each row is a different person and columns are questions. With that being said, we recently found out that one of the rows (which typically only includes the perspective of one person) actually represents multiple people so it would be ideal to treat this specific response/row differently. In a typically statistical program (like SPSS/Stata/R) i can simply ‘weight’ this response to equal the number of people it actually represents but unsure how to do this in Knime (e.g., what nodes to use). My workflow is still in development since it is unclear whether this issue can be addressed using Knime.

Overall, we would like to use text mining to analyze this information since the survey questions are all open-ended and 100’s of people have responded so far.

Hi @Cyrprano

Apologies, I am not a text mining expert, so I gotta ask a bit more to get an idea. What is in each cell? A number? The person’s response as text? How do you know which rows contains actually the answer of several people?

So from what I think the data looks like, I would probably go and filter the rows containing the answers of multiple people, and then use the Ungroup node to introduce a row (i.e. copies of that initial row) as many times as there are people, and then concatenate that back to the original table. So you end up with each row really corresponding to one person, thereby you omit the weighing.

I am positive you can do it in KNIME, but for sure it is a bit trickier to do than it might be in other programs.


1 Like

P.S.: in case you have a column that contains the information about how often you want to have a certain row, you can also use the One Row to Many node https://kni.me/n/D76rWzMcJyIID_ml :slight_smile:

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.