Retrieving, Organizing and Analyzing Text
See how to:
- Understand Text Explorer conventions and definitions related to analyzing unstructured text, including corpus, document, term, phrase, DTM (Document Term Matrix), tokenizing, stemming and stop word
See how to:
- Find the most common terms and phrases and determine the context in which terms or phrases are used
- Use term reports, phrase reports and word clouds
- Apply built-in and user-defined phrases
- Interactively customize, add or remove stop words
See how to:
- Find terms that tend to appear together
- Group and explore similar documents
- Uncover recurring themes (topics) within the collection of documents
- Cull important information from the text so it can be used in predictive models
- Use latent semantic analysis (SVD), scatterplot tendrils, topic analysis (Rotated SVD), Top Terms per Cluster Report and Term Probabilities by Cluster Report
- Save clusters or latent classes to get categorical predictors and save singular vectors to get continuous predictors
31+ minAdvancedMastering JMPText AnalysisVideoRe: Retrieving, Organizing and Analyzing TextThis is quite exhaustive! Thanks @brady_brady for this presentation, looking forward to reviewing this in relation to the more current posts in the community on this topic, such as: https://community.jmp.com/t5/Tutorials/Deriving-Sentiments-from-Opinions-or-Product-Choices/ta-p/317476.
At the very end in the last sectioned video coving the last sub-topic of:
- Save clusters or latent classes to get categorical predictors and save singular vectors to get continuous predictors
I think what would have been useful is walking through the example of launching a model on the continuous (e.g. singular vectors or topic vectors) and/or categorical predictors (e.g. latent classes or cluster numbers) vs the "Target Y" response. Like what kind of models you might use for this? PLS? Ordinary Logistic? Neural? (Which ones are typically the ones people use for these types of situations?)
It would be good to see and absorb (appreciate) an example dataset where we have a practically meaningful and reasonably measured Y response variable in the context of this type of data where it's mostly text information on X and maybe one (hopefully at least one) meaningful Y response.
Re: Retrieving, Organizing and Analyzing TextHi @PatrickGiuliano . One of Brady's @brady_brady colleagues, Kemal Oflus @kemal_oflus will be presenting his updated version of this topic at a June 10, 2021 Mastering JMP using JMP 16 Pro. It is a big topic, soKemal will aim to meet basic needs of people relatively new to text analysis. I am copying him so he can digest and possibly address your suggestions, if appropriate for the larger audience.
