Defining Tasks Automatically

Once you have uploaded your logs to your Timeline project, you can define tasks automatically during the task definition step. To do so, click the Show suggestion button in the Task Definition Editor. This will have the program automatically select suggested tasks based on detected repeating actions. This is done using the AI Suggestions feature. AI Suggestions is a machine learning algorithm that looks for repeating patterns in uploaded logs. Obtaining accurate results  can require significant computing time and needs to be based on a large dataset containing precise data.

IMPORTANT NOTE - SUCCESSFUL USE OF AI SUGGESTIONS: For the AI Suggestions feature to work as intended Logs are required to be as follows:

  1. A log should contain at least 100 events (i.e. any recorded user actions). For more information about supported events, see Supported Log Events.
  2. A recorded task type should repeat at least 5 times in a recorded event set. Each of these repeating fragments should have at least 5 similar events (which can occur in any order). Timeline Recording Services can optimally detect up to 10 different repeating tasks. The number of instances of each is not limited.
  3. If a log contains 100,000 or more events, processing it can take several hours. To allow users to continue to use Timeline while the analysis is being performed the automatic task definition is carried out in background mode. If required, you can suspend automatic task definition and continue defining them manually.

IMPORTANT NOTE – COLLECT ENOUGH TASK EXAMPLES. To sum up, for a task to be detected automatically, a user should carry out the corresponding activity at least 5 times, while the number of basic actions should be at least 100.

IMPORTANT NOTE – UNDERSTANDING EMPTY VARIANTS.

If there are too few task instances captured in an otherwise large dataset, it is possible that using the AI Suggestion algorithm will yield many inaccurate suggestions. In the UI, some suggestions can also be visualized as having 0 variants. This result indicates that the algorithm has found repetitive overlapping patterns that cannot be visually presented in the Task Definition UI. To proceed, these variants need to either be removed or analyzed in separate projects.

22.09.2023 8:59:47

Usage of Cookies. In order to optimize the website functionality and improve your online experience ABBYY uses cookies. You agree to the usage of cookies when you continue using this site. Further details can be found in our Privacy Notice.