Jump to: navigation, search

Model

The Genesys Content Analyzer classification model is a resource that Classification Server uses to classify email messages and other text objects. A model is associated with a category structure; it contains a statistical representation of each category in the structure. To classify a new message, Classification Server compares it with the representation of each category. The server then returns a list of categories, each of which has a percentage rating of the confidence with which the message can be assigned to that category. Model files are created and refined by training, and they are stored in the Universal Contact Server (UCS) database.



Glossary

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Category

In eServices, an item in a system of categories and subcategories, called a category tree, that is created and edited by using eServices Manager (previously, Knowledge Manager). Besides denoting a concept, a category may be associated with one or more standard responses, and one or more screening rules. If an incoming e-mail is assigned to the category, the category’s standard responses can serve either as the content for an automated reply or as suggestions for agents to use in their replies. A terminal or leaf category is one that contains no subcategories. A nonterminal category is one that does contain subcategories.
See also Knowledge Management.



Glossary

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Natural Language Processing

A Genesys Content Analyzer technology that operates on data that is in a language that is used by humans (as opposed to programming or other machine languages).



Glossary

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Content Analyzer Plug-in for GAX

Genesys Content Analyzer is an option to Genesys Knowledge Management, requiring an additional license. It adds natural language processing technology to Genesys Knowledge Management.

Content Analyzer Plug-in for GAX is an administration tool for managing Genesys Content Analyzer. The rest of Knowledge Management is handled in eServices Manager Plug-in for GAX.

Tip
  • Content Analyzer is designed to be viewed at a minimum screen resolution of 1280х1024, or a full HD resolution of 1920x1080.
  • For Content Analyzer to work with Microsoft Internet Explorer 11 you might have to adjust your browser settings. Ensure that the following Internet Explorer Security options are enabled:
    • Downloads/Font download
    • Scripting/Active scripting

What Does Content Analyzer Do, and How Does It Do It?

Content Analyzer looks at a text-based interaction and tells you where on your category tree it is most likely to belong.

  • It does this using a classification model, which is basically a statistical representation of a category tree.
  • It applies the model to an incoming text interaction (email, SMS, chat, and so on), and produces a list of categories. For each category, it tells you how likely it is that the interaction belongs to the category.
  • The actual analysis is done by Classification Server when it is triggered by a Classify or Analyze object in a routing strategy.
Tip
Classification, like screening, can operate on any interaction that has text somehow associated with it, whether as the body of the interaction (email, chat), or otherwise (as user data, for example). This document calls all of these things training interactions or messages.

So Where Do Models Come From?

First you need a Training Data Object. That is a category tree plus a set of text objects (such as emails, SMS messages, Twitter interactions, and so on), with each text object assigned to one category in the tree. Those text objects are called Training Messages or Training Interactions.

Then you run the training, and that produces the model.

In a little more detail,

  1. Create a Training Data Object.
  2. Modify the Training Data Object as needed:
    • Create some new Training Messages.
    • If you have uncategorized Training Messages, move them into categories.
    • Use cut/copy/paste to move Training Messages between categories.
  3. Get ready to run the training.
    • Choose when to run it and how to run it.
    • If you want, you can have it filter out confusing or irrelevant text strings before it runs.
  4. (coming soon!) Test the model.

Importing and Exporting

You can import and export training objects and models as part of the process of importing and exporting objects in eServices Manager. This is also a means of copying training objects.

Warning
Once you import an archive that includes any Training Data Object, you cannot import the archive again to a different language. If you attempt to do so, you will receive a message asking you to change the root category name, but when you do that you receive an error message saying that a Training Data Object with that name already exists.

As a workaround, you can rename the Training Data Object before importing the second time.
This page was last modified on December 11, 2018, at 11:49.

Feedback

Comment on this article:

blog comments powered by Disqus