Home > Managing Programs > Defining a New Program > Managing the Program Structure > Configuring Recognition Tasks

Configuring Recognition Tasks

By default, the linguistic recognition tasks of the Program are performed by SpeechMiner throughout the call and for any speaker. In other words, SpeechMiner searches for each item throughout the entire audio record of the call. You can specify exactly when a particular linguistic item should be sought in a call. Defining the search time makes searching more efficient and minimizes false positive identifications. In some systems, you can also specify that an item should only be sought in a call if a particular speaker or type of speaker (caller, agent, etc.) took part in the call.

Note: You can also specify other preconditions for linguistic items by defining and configuring triggers. See Configuring Trigger Properties.

Non-linguistic recognition tasks are performed on all calls and on all parts of each call. Silence, busy tomes, music and noise, dial tones, DTMF key presses, and ring-back tones are always detected. By default, non-verbal agitation is not sought. However, you can choose to search for it if you wish.

Configuring Linguistic Recognition Tasks

You can specify when each linguistic recognition task should be sought in a call by defining a time range in which it can appear. Both the start time and the end time of the range can be defined relative to either the start or end time of the call. If the recognition task is the child of another item in the Program structure, you can also define the times relative to the start or end time of the parent item.

In some systems, you can also specify that the recognition task should only be sought in the call if certain types of speakers took part in the call. For example, you could choose to perform the recognition task only if a supervisor participated in the call.  

To configure a linguistic recognition task:

  1. In the List pane, select the Content tab. The recognition tasks that are included in the Program are listed.
  2. Double-click the recognition task. The Recognition Task dialog box opens.

    Note: You can also open the dialog box by double-clicking the rectangle representing the task in the Program structure diagram.

    Recognition Task dialog box for linguistic tasks (Scripts)

  3. Under From, define the earliest time that the start of the task should be sought in the call audio, in seconds.

    For both options, select one of the following types of reference points:


  4. Under To, define the latest time that the start of the task should be sought in the call audio, in seconds, and select the reference point as explained above.
  5. In some systems, you can choose one or more types of speakers for the recognition task. If you can do so in your system, the Any checkbox is active. Clear the checkbox if you want to select one or more types of speakers from the list, and then select the speakers.
  6. Click OK. The setting is saved.

 Configuring Non-Linguistic Recognition Tasks

You can choose whether SpeechMiner should search for evidence of agitation in calls that belong to the Program.

To configure the non-linguistic recognition task:

  1. In the List pane, select the Content tab. The recognition tasks that are included in the Program are listed.
  2. Double-click the Non Linguistics task. The Recognition Task dialog box opens.

    Note: You can also open the dialog box by double-clicking the Non Linguistics rectangle in the Program structure diagram.

    Recognition Task dialog box for non-linguistic tasks

  3. If you want SpeechMiner to search for evidence of agitation in calls belonging to the Program, select Agitated. If you do not, clear the checkbox.
  4. Click OK. The setting is saved.


See also

Configuring Trigger Properties