Why Perform a Multivariate Search?
A Multivariate Search enables you to combine searches of faces, logos, objects, text, and/or transcription to quickly pinpoint relevant media.
This means you can simultaneously search against a variety of cognitive engines to locate media, which is especially helpful when searching through video.
How to Perform a Multivariate Search
1. To perform a Multivariate Search in the Discovery App, start by determining what media you would like to locate.
For example, you would like to find Mentions where Anderson Cooper appears on screen and “healthcare” is being discussed.
2. Click on the Search Bar. The default engine category selected is Transcription, so you can start by entering your keyword. For example, “healthcare”. Then click Add. Results in which the word “healthcare” is found in the transcription will populate:
3. To add the second search variate, click on a different engine category icon.
In this case, we would like to find Mentions where Anderson Cooper appears on screen, so we would select the Face Icon.
Begin typing the name of the face you are searching for, i.e. “Anderson Cooper”. As you type the name, trained faces* will populate in the menu below the Search Bar. Once your desired option appears, click on the face, then click Add to add it to the search:
*To search for a face or object entity, you will have to have trained the engine with that entity using Veritone's Library App.
4. The Search results will contain only Mentions where “healthcare” is found in the transcription and Anderson Cooper’s face appears on screen:
- Only engine categories that are contracted for your Organization will be available for multivariate search.
- Boolean Search can provide even more opportunities to pinpoint your search.
- To search for faces, logos or object that have not been trained in your available trained Libraries, contact your Customer Success Manager to discuss adding to your organization’s Library.