Business Context
A retail company receives thousands of customer reviews daily. They want to automatically classify these reviews into broad topics such as 'Shipping', 'Product Quality', and 'Billing' to route them to the correct department. This scenario tests the basic functionality of the action using standard English rules.
Data Preparation
Creation of a rule definition table for classifying feedback.
Copied!
| 1 | |
| 2 | DATA mycas.feedback_rules; |
| 3 | LENGTH config $200; |
| 4 | INFILE DATALINES dsd; |
| 5 | INPUT config $; |
| 6 | DATALINES; |
| 7 | "CATEGORY:Shipping,(OR, 'late', 'delay', 'not received', 'tracking')" "CATEGORY:Billing,(OR, 'charge', 'invoice', 'credit card', 'refund')" "CATEGORY:Quality,(OR, 'broken', 'damaged', 'poor', 'defective')" ; |
| 8 | |
| 9 | RUN; |
| 10 |
Étapes de réalisation
1
Load the rule definitions into memory.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | TABLE.loadTable / path='feedback_rules.sashdat' caslib='casuser' casout={name='feedback_rules', replace=true}; |
| 4 | |
| 5 | RUN; |
| 6 |
2
Compile the category model using default parameters (English).
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | textRuleDevelop.compileCategory / TABLE={name='feedback_rules'} config='config' casOut={name='feedback_model', replace=true}; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The action should successfully compile the text rules and generate a binary model table named 'feedback_model' (MCO file). This model can subsequently be used by the applyCategory action.