LDA, CTM and STM

This workflow is a simple implementation of topic models. Users only need to replace the excel reader node for their own data (e.g., a csv or xls file). The use of pre-processing is optional. Users can select additional text pre-processing nodes if needed.

In the last step users need to configure the Topic Extractor (LDA) node or Topic Extractor (STM) component. A node is different from a component in the sense that the component includes a series of nodes within the component. Please see the node and component descriptions for more details. The LDA node offers an additional measure for model fit (Goodness of Fit component).

The workflow is designed to provide a 2 topic solution, but this option can be changed by users so they identify what is the topic solution (K) that results in greatest model fit and interpretability

If you use this workflow, please cite: Villarroel Ordenes, Francisco, Grant Packard, Davide Proserpio, and Jochen Hartmann, “Using Text Analysis in Service Failure and Recovery: Theory, Workflows, and Models”, Journal of Service Research, Forthcoming.