Siebel Smart Answer Guide > Configuring and Implementing Siebel Smart Answer >

Creating and Training Your Knowledge Base Model

After you successfully install the Siebel Smart Answer Administration Tool, and create and categorize your corpus file, you are ready to create and train your knowledge base model.

To create and train your knowledge base model

1. Create and analyze your knowledge base model in the Siebel Smart Answer Administration Tool.

   NOTE:  If you want to create a hierarchical knowledge base, use the Category Editor to define the structure and use this category set to apply training and analysis on.

2. When you use the Siebel Smart Answer Administration Tool to create and analyze your knowledge base model, the main corpus is used to both create and evaluate the knowledge base model, as follows:
   • Training the corpus - the entire corpus is used to create the knowledge base model.
   • Testing the corpus - the corpus is randomly split 5 times into a testing and a training set. A knowledge base model is built and tested for each of the 5 random splits. The performance is averaged over the 5 tests and reported in the Benchmarking and Analysis Tab of the Siebel Smart Administration Tool.
3. Train and test the knowledge base model within the Siebel Smart Answer Administration Tool.

   In general, if you increase the number of items that you use for training the knowledge base model, the performance of the knowledge base model improves.

   Before you introduce your knowledge base model into a production environment, use the Knowledge Base Status, Benchmarking and Analysis workflow in the Siebel Smart Answer Administration Tool. For more information on using the Benchmark and Analyze Work Flow, see Creating a Knowledge Base Model Using the Benchmark and Analyze Work Flow.

4. Inspect the results and check the score for each category to verify that the results are accurate. As a final check you can optionally use the Query and Batch Query workflow in the Siebel Smart Answer Administration Tool detailed in Querying the knowledge base model to enter email messages either one at a time or in bulk in order to further verify the accuracy of the currently built knowledge base.

   After you begin using a knowledge base file to categorize incoming email and requests, and agents select categories or give feedback, the feedback is sent to the Siebel Smart Answer Server where it is stored for later processing.

5. Review the Siebel Smart Answer Administration Tool Benchmarking Analysis Tab to evaluate the training results.

   The categories are listed in a table at the top of the screen in descending order of performance. You can select a category and view the mis-categorized entries for that category in a table at the bottom of the screen. Each entry is editable, so that if you determine this entry is adding confusion to the knowledge base model, you can:

   • Re-label the entry with a category label (either existing or new).

     For more information on how to edit categories, see Modifying Categories.

   • Edit the content of the entry to make it more semantically informative.
   • Remove the entry.

     NOTE:  Each time you modify the contents of a category or edit the category set, you must repeat the Benchmarking and Analyze Work Flow in order to measure the impact of your changes.