Experiments allow you to select from available Cognitive Engines in your account and compare and evaluate metrics across your dataset.

To create an Experiment:

  1. Navigate to the desired Dataset you would like to evaluate

  2. Once you have opened the Dataset, select the Experiments Icon at the top of your page.

Upon clicking the Experiment Icon the New Experiment popup will open. The first step is to name your experiment and choose a folder location within your Data Center to store your data for future use.

Selecting Cognitive Engines

After you have named your experiment you will want to choose Cognitive Engines to compare and evaluate against your Dataset assets.

Choose a Capability

Select the Capability in the left-hand column to display all available Cognitive Engines you can choose from.

Note:

  • Only capabilities defined during the Dataset creation process will display for selecting and evaluation.

  • Only Cognitive Engines that can support your Dataset's data type will display for selection. As an example, if your dataset contains images the only Cognitive Engines that can successfully process this data type will display.

Choose and Configure Cognitive Engines:

Upon selecting the capabilities you wish to evaluate, you will need to select one or several Cognitive Engines for comparison.

Note:

Depending on how your account was configured you may have several available Cognitive Engines per capability or you may have no Cognitive Engines in a given category. If a capability is missing Cognitive Engines please reach out to your customer representative to help enable the engines that are right for your use case.

Alternatively, you can create your own Cognitive Engine using Veritone's Developer Tools CLICK HERE to learn more.

Click the Use this Engine button to select the engine for evaluation. If the engine requires additional settings these will appear below the engine card.

Once you have selected all Cognitive Engines for evaluation, select the Submit button to begin processing your Experiment.

You can check the progress of your Experiment by viewing the Status on the Experiment page within your Dataset. Once your Experiment is complete you can view the metrics from the Experiment Detail page.

Did this answer your question?