Forcing documents into categories

Forcing documents into and out of categories enables you to override the category definitions created by the automatic category building techniques without changing the actual category definition. You may find that although the document contains terms that are used to define a particular category, the document itself should not be in that category. In this case, you can force the document out of that category without having to remove the terms from the category definition.

Forcing is used in special cases where a document fits (or does not fit) a category, but for one reason or another (for example, it contains a particular term) is assigned (or not assigned) to that category. For example, this can occur when a respondent uses sarcasm in his or her response, such as "The pizza was great. I am sure everyone loves burnt, cold pizza." Let's suppose that you had a category called Pos: [<Food> + <Positive>] to capture positive opinions regarding the food that a restaurant serves, and this response is assigned to that category. In this case, you might want to force this response out of the category.

To force in our out of categories

From within the Data pane, select the document that you want to force into or out of a particular category.
From the menus, choose Categories > Force In or Categories > Force Out. A submenu displays the list of categories from which you can select.
Select the category to which or from which you want to force this document. If you have created many categories, some may not be visible in the submenu.

In this case, choose More at the bottom of the submenu. The All Categories dialog box opens, in which you can select the category and click OK to apply the change.
If you want to force the document into a new category, select Create Empty Category. A new category appears in the category tree using a generic name.

Whenever a category contains one or more forced documents, a pseudo-category called either Force In or Force Out is displayed below the category name in the tree.

To clear a forced state

From within the Data pane, select the document that you no longer want to force into or out of a category.
From the menus, choose Categories > Force In to force in, or choose Categories > Force Out to force out. The categories in which the document is forced out of or into are preceded by a check mark.
Select the category in the submenu that is checked and for which you want to remove the force. The check mark is removed and the document is no longer forced.

To clear all forced states

From within the Data pane, select a record containing a Force In or Force Out.
From the menus, choose Categories > Clear All > Force Ins or Categories > Clear All > Force Outs. The forced state on the documents is cleared and they are no longer forced into or out of the categories.

Note: This feature is only available if your source text contains a unique ID. If the source text doesn't have a unique ID, you can add a Derive node between the source document and the Text Mining node. This feature only has an impact when running an interactive session. When you deploy the category model for non-interactive scoring, this piece of information is not preserved or used since it is based on a document ID.