Maestro AI
Maestro AI is used to identify the relationship between terms and documents by analyzing the document content and the applied terms during the term training process. This enables you to apply trained terms to documents by selecting smart classification as the term application method when configuring terms settings for the following content sources: Teams & Groups, SharePoint Online, OneDrive, and Google Drive.
For enhanced accuracy, Opus supports a manual review process for terms applied through the smart classification method. Reviewers can verify the applied terms on the My task > Smart terms for review page, and approve them if correct, or reclassify the document with a new term if necessary.
Note the following:
- Maestro AI currently is available in the following data centers:
- Microsoft Azure Cloud Platform data centers: Australia Southeast (Victoria), East US (Virginia), and North Europe (Ireland)
- Google Cloud Platform (GCP) data centers: GCP for Australia Southeast (Sydney)
- If you have a trial subscription for Opus, you can train up to 5 terms.
Training can only be executed once per day, and you can find the last training time in the upper-right corner of the page. If you retrain terms on the following day, the previously collected properties, such as training status, accuracy, and training scope, will be cleared and then recollected.
Start Training Terms
To get started, you need to add 5 to 500 terms to the page. Click Add term, select terms, and then click Save. If some terms don’t require training, you can select the terms, click Remove, and then click OK in the confirmation message to remove the selected terms from the page.
Before training terms, make sure the following prerequisites are met. Otherwise, the training will fail.
- The training must include a minimum of 5 terms and a maximum of 500 terms.
- Each term must have already been applied to at least 20 documents from SharePoint Online or OneDrive (.txt, .pdf, .docx, .doc, .pptx, .ppt, .xlsx, .xls, .msg), or from Google Drive (Google Docs, Google Sheets, Google Slides), and the content of these documents must be different, as there are document duplicate checks during the term training process.
You can find the documents classified with a specific term by creating a **Classification** search condition along with a **Content** **source** search condition on the [Search](#missing-link) page. You can then open these documents to view their content directly from Opus.
Refer to the following steps to train terms:
-
Click Start training and click OK in the confirmation message.
-
A Start training job will start, and you can navigate to Job monitor to view the job details. This job is used to train terms.
*Note: If you add terms after the training has already started, the newly added terms will not be trained until you start another training.
-
After the Start training job is completed, a Train terms for smart classification job will start. This job is used to calculate the accuracy of each trained term.
More details:
- **Training** **status** – Indicates the training status of each term:
- **Not** **started** – Terms have not been trained yet.
- **Training** – Terms are currently being trained.
- **Trained** – Terms have already been trained. Terms of this status are ready for smart classification.
- **Pending** **approval** – This status will appear when you remove a term that is currently being trained. When the training is completed, you need to remove the term again.
- **Accuracy** – Indicates the classification accuracy of each term:
- **N/A** – There is no classification accuracy.
- **Low** – The classification accuracy is less than 50 points.
- **Acceptable** – The classification accuracy is between 50 and 70.
- **High** – The classification is larger than 70 points.
- **Training** **scope** – Indicates the training scope of each term. You can click the link to open the **Training** **scope** page to view the documents where this term has been applied.
- **Auto-apply** – If this setting is enabled, trained terms will be automatically applied to documents via the smart classification method, regardless of whether the manual review process is enabled.
Manage Training Scope
To define the method on how you would like to gather documents for training, complete the steps below:
*Note: Before setting a method to gather documents for training, make sure you have synchronized documents from the content sources to Opus for search and management.
-
On the Training scope page, click Method.
-
Select a method on how you would like to gather documents for training:
-
Automatic - Retrieve the latest 500 documents for each term – Once training begins, Opus will automatically retrieve up to 500 of the most recently classified documents for each added term. No manual document selection is required.
-
Manual - Select and manage documents myself – If you select this method, you need to manually add documents for each term you want to train.
Click the Add button to open a panel where documents classified with the terms you want to train are listed. Select the documents that you want to use for training and click Save. To remove a document from the training scope, select the document and click Remove. Click OK in the pop-up confirmation message.
-
Specific - Get documents from a designated location – If you select this method, select a content source and then specific the location from which to load documents. Once training begins, Opus will automatically retrieve the classified documents for each term from the specified location. No manual document selection is required.
-
Click Save to save the configurations.
View Training Report
The Training report page displays the SharePoint Online and OneDrive documents where AI has assessed a trained term to be applied to them.
More details:
- **File name** – Displays the name of the document. You can click the name to open the document directly from Opus.
- **Smart term** – Displays the term that Maestro AI has assessed to be applied to the document.
- **Current** **term** – Displays the term that is currently applied to the document.
> ***Note**: To ensure the column value accurately reflects the term applied to the document, it is essential to confirm that the **Content** **synchronization for** **search** jobs, which are used to synchronize content from SharePoint Online and OneDrive to Opus, have completed, otherwise, the term name displayed here might be incorrect.
- **Unique** **ID** **– Displays the unique ID of the document.
> ***Note**: Make sure the **Content** **synchronization for** **search** jobs has completed, otherwise, the unique ID might be empty.
- **Approval** **status** – Displays the following values:
- **Auto applied** – This value will be displayed when the **Auto-apply** setting is enabled for the trained term.
- **Waiting for approval** – This value will be displayed when the manual review process is involved, and the reviewer has not completed the review.
- **Approved** – This value will be displayed when the manual review process is involved, and the reviewer has approved the AI-assessed term.
- **Reclassified** – This value will be displayed when the manual review process is involved, and the reviewer disagree with the AI-assessed term, opting to reclassify the document with a new term.
- **Type** – Displays the file type.
- **Smart term assessed time** – Displays the time when the AI assessed the trained term to be applied to the document.
To export content, click Export, select a time period, and click Export. A download process will start in Download center. When the download process is completed, you can download the file by clicking Download.
