Deriving Sentiments from Opinions or Product Choices
See how to:
- Define basic text mining components: corpus (data table), document (row), Document Term Matrix (occurrences of terms of interest in corpus)
- Curate data using distribution, recode, stemming and Word Cloud to prepare for modeling
- Model the data using Fit Model, Generalized Regression, Elastic Net and BIC model validation
- Examine statistics, Solution Path and active parameters to evaluate model
Toronto Casino Study Shown in Video
- Problem:
- Data from a 2012 public poll on public desire for a proposed casino.
- ~18,000 Survey Responses
- Each Question has a score and a free text field.
- Use as Proof of Concept for DSA workflow.
- Findings:
- Sentiment regarding the casino project generally falls along ideological lines.
- Those who feel the good of the city should take priority are in favor.
- Those who feel the good of the people themselves should take priority are generally against.
- Sentiment Analysis is possible using the DSA workflow.
The Rains of Castamere Problem Study Shown in Video
- Problem:
- Analyze the data from the twitter stream around the Game of Thrones Episode, “The Rains of Castamere.”
- Develop a sentiment model from the twitter stream.
- Challenge:
- Need to use the emoji in the tweets themselves as the response in the model.
- Need to extract , decode, and consolidate the emoji into something to model
- Findings:
- A script was employed to extract the emoji and recode them into something understandable.
- Using Factor Analysis, the underlying “sentiments” could be extracted and correlated to emoji that are used to express them.
Resources
31+ minIntermediateMastering JMPText AnalysisVideo
