on recommondation of my Professor I’m using Knime for my master thesis. I try to analyse customer reviews from a Yelp dataset on the influence of Service quality on Customer satisfaction. It was also recomended to use SC-LDA, which is based on a Paper: Büschken, J., & Allenby, G. M. (2016). Sentence-based text analysis for customer reviews. Marketing Science , 35(6), 953-975. First I taught to just split the reviews into single sentences but in after reading the paper completely I saw that there is more to it than just that.
Now to my question: I’m not quiet shure what this adaption changes and how or if i can add this into knime? Can somebody help me with that?
At the time being, the default node (Parallel LDA) is the original algorithm. There are many versions of LDAs in academic journals, one of them is as you pointed out in that paper.
If you are using Knime for LDA, you either have to deploy the original algorithm (meaning you cant use the referenced paper anymore), or develop your own Node that uses different algorithms (which requires various knowledge out of you) – asking people to do that for you is actually a very big ask
…
On another related topic, you mentioned two constructs for your research. The way your wording is phrased tells me that your research design is not exploratory in nature. You might want to re-discuss with your professor as to why they think LDA (which is exploratory in its nature) is the suitable method for what appears as either confirmatory or explanatory research. In general, a research method is determined by research questions, and whether your constructs already have pre-established variables from academic publications. Decisions on the research method (and the tools used to carry it out) usually comes afterward, not the other way around.
