Note that these instructions are for setting up a new account on Textual Cloud. For a self-hosted instance, depending on how it is set up, you might either create an account manually or use single sign-on (SSO).

Signing up for Textual

To get started with a new Textual account:

1. Go to https://textual.tonic.ai/.

2. Click Sign up.

3. Enter your email address.

4. Create and confirm a password for your Textual account.

5. Click Sign Up.

Textual creates your account and sends you an email message to activate the account.

After you activate your account and log in, Textual displays the Textual home page, which you can use to preview how Textual detects and replaces values. For more information, go to .

When you set up an account on Textual Cloud, you start a Textual free trial.

On the home page, the Getting Started section provides links to tasks and information related to:

• Using the

• Working with

• Working with

• Working with

To view the available options for a panel, hover the mouse over it. Hover over an option to display a tooltip to identify the action that the option performs.

During the free trial Textual scans up to 100,000 words for free. Note that Textual counts actual words, not tokens. For example, "Hello, my name is John Smith." counts as six words.

After the 100,000 words, Textual disables scanning for your account. Until you purchase a pay-as-you-go subscription, you cannot add files to a dataset.

Your current usage is displayed in the navigation menu.

Textual also prompts you to purchase a , which allows an unlimited number of words scanned for a flat rate per 1,000 words.

You can also request a Textual product demo.

Textual billing is based on the number of words that it scans to identify sensitive values, called entities.

Billing is per 1,000 words

Textual always bills for every 1,000 scanned words. Textual counts actual words, not tokens. For example, "Hello, my name is John Smith." counts as six words.

Free trial versus pay-as-you-go versus contract accounts

A free trial account is not billed, but is restricted to a specific number of words. After they reach that limit, Textual no longer performs any scans and does not allow the account to add files to datasets.

A pay-as-you-go account does not have a limit. The account credit card is billed each month based on the number of scanned words from the previous month.

A contracted account does not have a limit, but it might have a number of words that it pays for when the contract starts. If the account exceeds that word count before the contract period ends, then they are billed for the additional scanned words.

What counts as a scan?

A scan specifically means that Textual analyzes a piece of text or a file to identify sensitive values.

This includes when:

• You add a new file to a dataset. Textual immediately scans the file.

• You change the dataset configuration in a way that requires Textual to scan the files again. For example, you change the activated custom entity types. Textual prompts you when a new scan is required.

• You use the SDK or API to send redaction requests for individual text strings or files.

Textual does not bill you for the following:

• Failed scans of a file in a dataset or an SDK request. For example, if Textual is unable to parse a file, then it does not count that file as having been scanned.

• Failed SDK requests to redact text, JSON, XML, or HTML strings. If the response is not 200, then the text is not considered scanned.

• New rendering of a dataset output file. For example, when you change the entity type handling or synthesis configuration, Textual does not re-scan the file.

Tonic Textual allows you to put your text-based data to work for you.

A Textual dataset is a collection of files from a local file system or cloud storage. Textual scans the dataset files to identify sensitive values. You can then choose to redact or replace those sensitive values, to produce output files in the same format that you can safely use for development and training.

A guided redaction project identifies blocks out sensitive values in files. For example, you might use guided redaction to prepare documents in response to a Freedom of Information Act request. Textual scans the files to identify values. You can then review and adjust the results before you download the output files.

You can use the Textual Python SDK or the Textual REST API to manage datasets or to remove sensitive values from individual text strings and files.

Want to know what's in the latest Textual releases? Go to the .

The Textual Agent is a chat-based tool to help you to explore the content of dataset files and configure the file processing.

The Textual Agent starts when you create or edit a Textual dataset.

Each combination of dataset and user has its own Textual Agent chat.

The Textual Agent is always enabled on Textual Cloud.

For information on how to enable the Textual Agent on a self-hosted instance, go to .

On the Entity Types page, the Progression Tracker column indicates the completion status of each step in the process.

For each step, the block color indicates the status:

• Grey - The step is not started.

• Yellow - The step is in progress.

To start a new custom entity type:

1. On the Additional Entity Types page, click Add Custom Entity Type.

2. On the Add Custom Entity Type panel, click Model-based, then click Next.

Before you can , you must select the model to use.

On the model list page, the published model is marked as Active.

To change the active version:

1. Hover the mouse over the model to make the active model for the entity type.

2. Click Activate.

To change the entity type name, on the entity type details page:

1. Click the actions menu next to the entity type name.

1. In the menu, click Edit entity name.

To delete a dataset:

1. Either:

   1. On the Datasets page, click the delete icon for the dataset.

   2. On the dataset details page, click

The dataset name displays at the top left of the dataset details page.

To change the dataset name:

1. On the dataset details page, click Project settings.

2. On the Dataset settings page, in the Dataset Name field, provide the new name for the dataset.

Tonic Textual can process the following types of files:

• .txt

• .csv

• .tsv

A detected value might match multiple entity types.

For example, a telephone number might match both the Phone Number and Numeric Value entity types.

On most dataset details views, each value is only counted once, for the entity type that it is assigned in the output file. The Analytics page under Entities analysis allows you to choose whether to include a value in the counts for all of the types that it matches.

By default, a detected value is assigned the entity type that it most closely matches. For our example, the telephone number value most closely matches the Phone Number entity type, and so by default is included in the Phone Number count and values list.

If the entity type is ignored, or the value is excluded, then Textual moves the value to the next matching type.

In our example, if you set the handling option for Phone Number to Ignore, then the telephone number value is added to the count and values list for the Numeric Value entity type.

The Project files and Entity settings page display the following summary results for the dataset::

• The number of detected entities

• The percentage of dataset content that is sensitive

For each dataset file, the Project files page displays:

• The number of detected entity values in the file. The count does not include values for entity types for which the entity type handling is set to Ignore.

• The total number of words in the file.

If you also use Tonic Structural, then you can configure Textual to enable selected synthesized values to be consistent between the two applications.

For example, a given source telephone number can produce the same replacement telephone number in both Structural and Textual.

To enable this consistency, you configure a statistics seed value as the value of the Textual SOLAR_STATISTICS_SEED. A statistics seed is a signed 32-bit integer.

The value must match a , either:

• The value of the Structural environment setting TONIC_STATISTICS_SEED

To display the synthesis options for an entity type, click the icon next to the handling option dropdown list.

For Email Address values, you can choose to preserve the original email address domains. For example, source email addresses for example.com remain example.com email addresses in the output.

From the dataset file list, to display the preview, either:

• Click the file name.

• Click the options menu, then click Preview.

For de-identified text files, the value highlighting color indicates the entity type handling option.

• Redacted values are purple.

• Synthesized values are blue.

• Ignored values are grey.

From the file preview, to select the for an entity type:

1. On either de-identified or original view, click a detected value.

2. On the details panel, click the entity type handling option. Textual applies the same option to all entity values of that type.

For the Synthesis option, you can configure a specific replacement value for the original value. For example, for the original first name value John, you can specify to replace all instances with Michael.

From a PDF file preview, you can choose to ignore a specific value:

1. In de-identified or original view, click the value.

2. On the details panel, to ignore the value, toggle Ignore this instance to the on position.

For a dataset that generates JSON output, the file preview displays the original file content.

For a text file, you can switch between Markdown and the JSON output.

For CSV and RTF files, you can display either Markdown, HTML, or JSON output.

For a PDF or file, you can always display Markdown, HTML output, or JSON output.

If the file contains tables, then the Tables option displays, to allow you to view the tables specifically.

If the file contains key-value pairs, then the Key-Values option displays, to allow you to view the key-value pairs specifically.

To generate original format output files for a cloud storage dataset, on the dataset details page, click Generate to <cloud storage type>.

Tonic Textual generates the output files to the . If the output location is not configured, then the generate option is disabled.

For datasets that produce JSON output, Textual generates the output files automatically as soon as the output location is configured.

To create a project:

1. On the Guided Redaction page, click New Project.

2. On the new project panel, in the Project Name field, provide a name for the project.

3. Click Save.

Textual displays the project details page.

You can assign tags to each guided redaction project, to help to further identify and link projects. For example, you might use a tag to indicate when a project is for a FOIA request.

On the Guided Redaction page, the Tags column contains the assigned tags for the project.

On the project settings panel, the Tags field lists the project tags. To display the settings panel, either:

• On the Guided Redaction page, click the settings icon for the project.

• On the project details page, click More, then click Settings.

To change the assigned tags:

1. Click the Tags column or field.

2. To add a tag, type the tag text, then press enter.

3. To remove a tag, click its delete icon.

4. To remove all of the tags, click the delete icon at the right of the tags field.

The User Profile page displays a summary of information about your Textual account.

From the user profile page, you can copy your organization identifier, configure a team name, and manage your Textual API keys. For more information about managing your Textual API keys, go to Creating and revoking Textual API keys.

Displaying the User Profile page

To display the User Profile page:

1. Click the user icon at the bottom of the navigation menu.

2. In the user menu, click User Profile.

The profile summary includes the identifier of your organization in Textual.

To copy the identifier, click the copy icon.

From the User Profile page, you can specify a team name. For example, you might use the team name field to identify a specific department or project that you belong to.

To add a team name, type the name in the field.

You can ask the Textual Agent to provide information about the content of the dataset files and to update the entity handling configuration.

The following are just a few examples:

An entity type is a category of entity value. For example, the entity value John might be an example of the Given Name entity type.

Tonic Textual comes with a built-in set of entity types that are detected by default. These include the more common entity types such as names and locations.

Textual also provides a library of additional types that you can import into Textual and use in your datasets.

You can also configure custom entity types, which you can use to detect values that are not covered by the built-in entity types or the types in the library.

When you create a custom entity type, you can either:

• Use regular expressions to identify matching values. You might use this option when there are a limited number of values, or if the values follow specific formats that can easily be identified with a regular expression.

• Training a model is an iterative process that can take hours or days, depending on your data. You might create this type of custom entity when there are a large number of values that do not follow a specific format. The values need to be identified more by context.

You can also view this .

The Entity Types list displays the additional types that you created or imported, and provides access to the list of built-in types.

To display the Entity Types page, in the navigation menu, click the entity types icon.

Information in the list

For each entity type, the Entity Types page includes the following information:

• Name of the entity type.

• For added model-based types:

  • The progression tracker. For more information, go to .

  • The published model.

• Whether the type is an added regex type, an added model-based type, or a type that you imported from the entity type library.

• Whether the type is enabled for any datasets and if so, the number of datasets that it is enabled for. For more information, go to .

• When the type configuration was most recently updated.

You can filter the list by type name, by category, and by the user who created the type.

To filter the list by the entity type name, in the search field, begin to type text in the name. As you type, Textual filters the list to only include matching entity types.

To filter the list based on whether the entity is regex-based, model-based, or was imported from the library:

1. Click Categories.

2. Check the checkbox next to each source type to include.

To filter the list to only include custom entity types that were created by specific users:

1. Click Creator.

2. In the dropdown list, check the checkbox for each user to include.

The entity types are always grouped by type. By default, within each type, they are sorted by name.

To reverse the order of the groups and names, click the Name column heading.

To sort by when the types were most recently updated, click the Last Updated column heading. To reverse the sort order, click the column again.

You can export a model-based entity type to an encrypted file. The export includes the entire configuration.

You can then import the entity type into another instance of Textual.

Exporting an entity type

From the details page for an entity type, to export the entity type:

1. Click the actions menu next to the entity type name, then click Export Entity Type.

1. After it downloads the custom entity type to the encrypted file, Textual displays the encryption key that was used to encrypt the file, and that you will need when you import the entity type. To copy the key, click the copy icon.

After you export a model-based custom entity type, you can import it into another instance of Textual.

When you import an entity type, you are prompted to provide the encryption key that Textual provided during the import.

You cannot use the import to update or replace an existing entity type. If there is already an entity type with the same name, the import fails.

For an imported type, the progression tracker reflects the status of the type. However, regardless of whether the imported entity type was complete and had a published model, it is does not have a published model when it is imported. After the import, you can update the entity type to identify a published model.

To import an exported entity type:

1. On the Custom Entity Types page, click Import type, then select Exported custom model.

2. On the Import Entity panel, to search for and select the exported file, click Choose file.

1. In the Decryption key field, paste the encryption key that you copied during the export.

2. Click Import.

In a dataset, Textual automatically scans for the built-in entity types. For other entity types, including custom types and imported types, you must first enable the entity type for the dataset.

Before you remove an entity type, you must make sure that it is not enabled for any datasets.

Viewing the dataset count for entity types

On the Entity Types page, the Activated for column displays the number of datasets that the entity type is enabled for.

Enabling and disabling an entity type for datasets

From the Activated for column, you can view and change the selected datasets for the entity type.

You can also enable or disable entity types from the dataset details. For more information, go to Working with additional entity types.

You cannot select datasets for model-based entity types that do not have a published model.

From the Activated for column on the Entity Types page, to change the selected datasets:

1. Click the database icon. The Activate custom entity panel displays the datasets that you have access to.

1. To filter the datasets by name, in the search field, type text from the name.

2. For each dataset, the toggle indicates whether the entity type is active for the dataset. When the toggle is in the on position, the entity type is enabled for the dataset. To enable or disable the entity type for a single dataset, set the toggle.

3. When you enable the entity type for a dataset, you can also set how Textual prioritizes the custom entity type when a detected value matches both the custom type and a built-in type. The options are:

1. To enable or disable the entity type for all of the datasets that are currently included in the list, click Bulk Edit, then:

   • To enable the entity type for all of the datasets, and always prioritize the custom type over built-in types, select Enable (Prioritized).

   • To enable the type for all of the datasets, and use the confidence scores to determine which type takes precedence, select Enable (Standard).

Tags can help you to organize your datasets. For example, you can use tags to indicate datasets that belong to different groups, or that deal with specific areas of your data.

You can manage tags from the Datasets page or from the Project Settings tab on the dataset details.

Datasets page

On the Datasets page, the Tags column displays the currently assigned tags.

On the dataset

To change the tag assignment for a dataset:

1. Click Tags.

2. On the dataset tags panel, to create a new tag, type the tag text, then press Enter. You can also select an existing tag from the dropdown list.

1. To remove a tag, click its delete icon.

2. To remove all of the tags, click the delete all icon.

On the Dataset settings page of the dataset details, the Tags section lists the currently assigned tags.

To change the tag assignment for the dataset:

1. Click the Tags field.

2. To create a new tag, type the tag text, then press Enter. You can also select an existing tag from the dropdown list.

1. To remove a tag, click its delete icon.

2. To remove all of the tags, click the delete all icon.

Tonic Textual uses dataset permission sets for role-based access (RBAC) of each dataset.

A dataset permission set is a set of dataset permissions. Each permission provides access to a specific dataset feature or function.

Textual provides built-in dataset permission sets. Organizations can also configure custom permission sets.

To share dataset access, you assign dataset permission sets to users and to SSO groups, if you use SSO to manage Textual users. Before you assign a dataset permission set to an SSO group, make sure that you are aware of who is in the group. The permissions that are granted to an SSO group automatically are granted to all of the users in the group.

To change the current access to the dataset:

1. Either:

   • On the Datasets page, click the share icon.

   • On the Dataset settings page of the dataset details, click Share.

1. The dataset access panel contains the current list of users and groups who have access to the dataset, and displays their assigned dataset permission sets. To add a user or group to the list of users and groups:

   1. In the search field, begin to type the user email address or group name.

   2. From the list of matching users or groups, select the user or group to add.

For a local file dataset, you upload and remove new files directly.

On Tonic Textual Cloud, and by default for self-hosted instances, Textual stores the uploaded files in the application database.

On a self-hosted instance, you can instead configure an S3 bucket where Textual stores the files. In the S3 bucket, the files are stored in a folder that is named for the dataset identifier.

For more information, go to Setting the S3 bucket for file uploads and redactions.

For an example of an IAM role with the required permissions, go to .

Adding files to the dataset

From the dataset details page, to add files to the dataset:

1. In the left menu, click Project files.

2. On the dataset files page, click Upload Files.

1. Search for and select the files.

You can also drag and drop files to the page.

Textual uploads and then processes the files. For more information about file processing, go to .

On a self-hosted instance, when a file fails to upload, you can download the associated logs. To download the logs, click the options menu for the file, then select Download Logs.

To remove a file from the dataset:

1. In the file list, click the options menu for the file.

2. In the options menu, click Delete File.

To display the list of files for the dataset, on the dataset details page, click Project files.

For cloud storage dataset, the file list displays the full path to each file.

Information in the file list

For each file, the file list includes the following information:

• The name of the file. For cloud storage datasets, the file name includes the full path to the file. When the file needs to be rescanned, a warning icon displays in front of the name.

• The number of detected entity values in the file. Note that Textual does not display an entity value count for CSV files.

• The total number of words in the file. Note that Textual does not display the word count for CSV files.

• When the file was added to the dataset.

• When the file was most recently scanned.

• For PDF and image files on self-hosted instances, the OCR model used to process the file.

To filter the list based on the file name, in the search field, begin to type text from the file name.

As you type, Textual updates the file list to only include matching files.

You can also sort the list by any of the columns.

• To sort by a column, click the column heading.

• To reverse the sort order, click the column heading again.

When you add files to a local files dataset, or change the file selection for a cloud storage dataset, Tonic Textual automatically scans the files to identify the entities that they contain.

When you change the dataset configuration, Textual also prompts you to run a new scan. For example, a new scan is required when you:

• Configure added values

• Change the available custom entity types

The file list reflects the current scanning status for the file. A file is initially queued for scanning. When the scan starts, the status changes to scanning. When Textual finishes processing a file, it marks the file as scanned.

When a file needs to be rescanned, a warning icon displays in front of the file name.

As Textual processes each file, it updates the results.

If needed, you can pause the file processing. To pause the processing, click the pause icon.

The information in the results only reflect the files that are scanned.

For a cloud storage dataset, when you generate output, Textual only includes files that are scanned.

After you pause the scan, you can start a scan on individual files.

To start a scan on a file, click the refresh icon for the file.

When Textual is unable to process a file, it displays an error for that file.

To download log files for the failed file:

1. Click the options menu for the file.

2. Click Download Logs.

The dataset details page provides information about the results of the sensitivity detection, including the overall results, the results per file, and the results per entity type.

You can also use the Textual Agent to explore the dataset results.

The Entities Catalog page displays the list of detected entity values for the dataset. To display the Catalog, in the left menu on the dataset details page, click Catalog.

Information in the Entities Catalog

The Entities Catalog lists each instance of an entity value separately. For example, the given name John is detected twice in one file and 3 times in another file. The Entities Catalog then contains 5 entries for John.

The Entities Catalog does not include entities for entity types that have the entity type handling option set to Ignore.

For each value instance, the Entities Catalog includes:

• The entity value.

• How the value appears in the output, based on the selected handling option for the value's entity type.

• The entity type.

• A confidence score to indicate how confident Textual is that the value is correctly detected and identified.

• The entity value instance in its immediate context.

• The name of the file that contains the value instance.

To filter the list by text in the entity value, in the search field, begin to type the text.

As you type, Textual filters the list to only include entity values that contain that text.

By default, the Entities Catalog list includes all of the entity types. To filter the list to a specific entity type, click All types, then select the entity type. To remove the filter, select All types.

By default, the Entities Catalog list includes values from all of the files. To filter the list to only include values detected in a specific file, click All files, then select the file. To remove the filter, select All files.

You can sort the Entities Catalog by the value, transformation, entity type, and confidence score.

To sort by a column, click the column heading.

To reverse the sort order, click the column heading again.

The Entities Analytics page displays a summary of the detected entity types. For each entity type, you can display the distribution across the dataset files.

It does not include entity types that have their entity type handling option set to Ignore.

You can also use the Textual Agent to ask questions about the detected entity types and counts.

Selecting the value count option

On the Entities Analytics page, you can choose how Textual determines the displayed value counts and entity types.

• Match counts to redacted files - Displays value counts based on the output files. For this view, the counts do not include entity types that are ignored. The counts also resolve entity values that match multiple types and entity values that share some text. Each value is counted as a single type.

• Show all detected entities - Displays the full detection value counts. For this view, the counts per entity type include all of the entities that Textual found during processing. This includes:

  • Values for ignored entity types

  • Entity values that match multiple entity types

The panels at the top of the page provide summary information for the detected entity values. The displayed values are based on the selected value count option.

The summary information includes:

• The number of detected entity values.

• The number of detected entity types.

• The percentage of detected values that are redacted.

• The percentage of detected values that are synthesized.

The entity types list on the Entities Analytics page displays a summary of the detected value counts for the detected entity types. The displayed entity types and counts are based on the selected value count option.

For each entity type, the list includes:

• The count of detected values

• The percentage of detected values in the dataset that are of that type

By default, the entity types are listed in descending order based on the value count.

You can sort the list by the entity type, count, and percentage. To sort by a column, click the heading. To reverse the sort order, click the heading again.

When you click an entity type, Textual displays a panel that lists the 10 files that contain the most detected values for that entity type.

The panel also allows you to change the handling option for the entity type.

The Entities Catalog lists the values that Textual detected in the dataset files.

The Dataset search allows you to search for specific values to determine whether Textual detected them and assigned the correct entity type.

You can also use the Textual Agent to ask whether the dataset includes specific values.

Starting a search

To start a search:

1. On the dataset details page, click Dataset search.

2. In the search field, provide the text to search for.

When the text matches a full word in a file, Textual displays the results.

Textual only searches for full words. For example, if you type "nu", it does not find the word "number". You must type the full word "number".

The search results include a separate row for each found instance of the search text.

For each instance, the results include:

• The instance itself, with the immediate context. The context includes the few words before and after the search text.

• The assigned entity type, if Textual detected the search text as an entity. If it did not, then the entity type is None.

• The name of the file where the instance was found.

At the top of the list, Textual displays the approximate number of matches and number of files that contain matches.

Depending on the number of matches, the list might not initially include all of the results. If there are additional results, then the Load more link displays. When you click Load more, Textual adds the next batch of results to the end of the list.

You can filter the results to only include matches in selected files.

To filter the results:

1. Click Filter by file.

2. In the file list, check the checkbox for each file to include the results for.

The Entity settings page displays the list of active entity types for the dataset set. This includes:

• All of the built-in entity types

• Any custom entity types that are active for the dataset

You can also use the Textual Agent to ask about the detected entity types.

Information in the entity types list

For each entity type, the list includes:

• The name of the entity type.

• The number of detected values for that type in the dataset files. For ignored entity types, the detected value count is 0.

• The selected handling option.

The list is sorted in descending order by the number of detected values. The entity type with the highest number of detected values is at the top of the list.

You can filter the entity types list by:

• Text in the type name or description.

• Whether the entity type is built-in or custom.

• Whether there are detected entities for the entity type.

• The handling option for the entity type

To filter by name or description, in the search field, begin to type text in the name or description. As you type, Textual filters the list to only include matching entity types.

To apply other filters, click Filter options, then select the filters to apply.

When Tonic Textual generates replacement values, those values are always consistent within the dataset. Consistency means that the same original value always produces the same replacement value.

You can configure how Textual generates the replacement values for each entity type.

You can also configure synthesis options from the Textual Agent.

For all entity types, you can:

• Map original values to replacement values.

• Provide a constant value to use to replace all of the original values, except for mapped values.

Mapping original values to replacement values

You can map specific original values to specific replacement values.

For example, for the Given Name entity type, you might indicate to always replace John with Michael and Mary with Melissa.

For original values that are not in the mapping:

• If you also provid a constant value, then Textual uses that value.

• Otherwise, Textual selects the values.

In the text area, provide a JSON object that maps the original values to the replacement values. For example:

With the above configuration for the Language entity type:

• All instances of French are changed to German.

• All instances of English are changed to Japanese.

The values are case-insensitive.

For example, if you specify "John": "Michael", then Structural also replaces john with michael and JOHN with MICHAEL.

Structural ignores leading and trailing punctuation.

To continue the example of "John": "Michael", Structural also replaces 'John' with 'Michael'.

You can provide a single constant value to use to replace all instances of the entity type. For example, you might replace all bank numbers with "0000000".

In the Constant Value field, provide the value to use.

When you provide a constant value, Textual ignores other .

However, Textual does respect any .

The following built-in entity types have additional type-specific synthesis options.

Before it synthesizes a name value, Textual always removes any leading or trailing punctuation, except for periods after initials.

For the Given Name and Family Name entity types, you can configure:

• Whether to treat the same name with different casing as a different value.

• Whether to replicate the gender of the original value.

Differentiating source values by case

To treat the same name with different casing as different source values, check Is Consistency Case Sensitive.

For example, when this is checked, john and John are treated as different names, and can have different replacement values - john might be replaced with michael, and John might be replaced with Stephen.

When this is not checked, then john and John are treated as the same source value, and get the same replacement.

To replace source names with a names that have the same gender, check Preserve Gender.

For example, when this is checked, John might be replaced with Michael, since they are both traditionally male names. However, John would not be replaced with Mary, which is traditionally a female name.

Location values include the following types:

• Location

• Location Address

• Location State

• Location Zip

For location types that include zip codes, you can also specify how to generate the new zip code values.

Under Address generator type, select the type of address generator to use:

• HIPAA-compliant address generator. This option generates values similar to those generated by the .

• Non-HIPAA address generator. This option generates values similar to those generated by the .

If you configured a Textual statistics seed that matches a Structural statistics seed, then the generated address values are consistent with values generated in Structural. A given address value produces the same output value in both applications.

For example, in both Textual and Structural, a source address value 123 Main Street might be replaced with 234 Oak Avenue.

For Location State, based on HIPAA guidelines, both the Synthesize option and the Ignore option pass through the value.

For location types other than Location State, by default, Textual replaces a location value with a realistic corresponding value. For example, "Main Street" might be replaced with "Fourth Avenue".

To instead scramble the values, uncheck Replace with realistic values.

By default, to generate a new zip code, Textual selects a real zip code that starts with the same three digits as the original zip code. For a low population area, Textual instead selects a random zip code from within the United States.

To instead replace the last two digits of the zip code with zeros, check Replace zeroes for zip codes. For a low population area, Textual instead replaces all of the digits in the zip code with zeros.

This option is also used when you truncate zip codes to the first 3 digits. When you truncate zip codes, and enable this option, Textual adds zeroes to fill out the zip code to 5 digits.

To truncate zip codes to the first 3 digits, check Use 3-digit zip codes.

If you also check Replace zeroes for zip codes, then Textual replaces the truncated digits with zeroes.

For zip codes from outside of the United States, then to replace those zip codes with zeroes, check Replace foreign zip codes with zeroes.

When you select the Synthesis option for the Age entity type, you can either:

• Shift the age within a specific range.

• Pass through ages younger than 90, and group ages 90 and older into a single value.

Shifting the age within a range

To shift the age within a specified range:

1. From the Age generator type dropdown list, select Age shift.

2. In the Range of Years +/- for the Shifted Age field, enter the number of years before and after the original value to use as the range for the synthesized value.

3. To shift all ages in a file by the same amount, check Apply same shift for entire document.

By default, Textual shifts the age value to a value that is within 7 years before or after the original value.

Based on the original age value, to either pass through or group the value, from the Age generator type dropdown list, select Passthrough or group age.

With this option:

• If the original age value is younger than 90, then the value is left as is in the output.

• If the original age is 90 or older, then the value is replaced with 90+.

By default, Textual scrambles age values that it cannot parse.

To instead pass through the value unchanged, uncheck Scramble Unrecognized Ages.

For Phone Number values, you can choose whether to generate a realistic phone number. If you do, then the generated values can be consistent with values generated in Structural.

Selecting the generator type

From the Phone number generator type dropdown list:

• To replace each phone number with a randomly generated number, select Random Number.

• To generate a realistic telephone number, select US Phone Number. The US Phone Number option generates values similar to those generated by the .

If you also configured a Textual statistics seed that matches a Structural statistics seed, then the synthesized values are consistent with values generated in Structural. A given source telephone number produces the same output telephone number in both applications.

For example, in both Textual and Structural, 123-456-6789 might be replaced with 154-567-8901.

The Replace invalid numbers with valid numbers checkbox determines how Textual handles invalid telephone numbers in the data.

To replace the invalid with valid telephone numbers, check the checkbox.

If you do not check the checkbox, then Textual randomly replaces the numeric characters.

The Preserve US area code checkbox determines whether Textual preserves the area code for telephone numbers in the United States.

To preserve the area codes from the source values in the output values, check the checkbox.

If you do not check the checkbox, then Textual replaces the entire telephone number, including the area code.

When you select the Synthesize option for Date/Time values, you can either:

• Shift the date within a specific range.

• Truncate all dates to January 1. The year is set based on whether the original year is more than 89 years ago.

You can also configure how to handle dates that Textual does not recognize, and add additional date formats for Textual to recognize.

Shifting the date within a range

To shift the date within a specified range:

1. From the Generator type dropdown list, select Date shift generator.

2. In the Left bound on # of Days To Shift field, enter the number of days before the original date within which the replacement datetime value must occur. For example, if you enter 10, then the replacement datetime value cannot occur earlier than 10 days before the original value.

3. In the Right bound on # of Days To Shift field, enter the number of days after the original date within which the replacement datetime value must occur. For example, if you enter

To truncate the date, from the Generator Type dropdown list, select Date Truncation Generator.

With this option, in the output:

• The month and day are always January 1.

• If the original year is less than 90 years ago, then the year remains as is in the output.

• If the original year is 90 years or more ago, then the year is set to the current year minus 89.

Textual can parse datetime values that use either a format in or a format that you add.

The Scramble Unrecognized Dates checkbox indicates how Textual should handle datetime values that it does not recognize.

By default, the checkbox is checked, and Textual scrambles those values.

To instead pass through the values without changing them, uncheck Scramble Unrecognized Dates.

By default, Textual is able to recognize datetime values that use a format from .

Under Additional Date Formats, you can add other datetime formats that you know are present in your data.

The formats must use a .

To add a format, type the format in the field, then click +.

To remove a format, click its delete icon.

When you select the Synthesis option for a Date of Birth, you can either:

• Shift the date within a specific range.

• Truncate all dates to January 1. The year is set based on whether the original year is more than 89 years ago.

You can also configure how to handle dates that Textual does not recognize, and add additional date formats for Textual to recognize.

Shifting the date within a range

To shift the date within a specified range:

1. From the Generator Type dropdown list, select Date shift generator.

2. In the Left bound on # of Days To Shift field, enter the number of days before the original date within which the replacement datetime value must occur. For example, if you enter 10, then the replacement datetime value cannot occur earlier than 10 days before the original value.

3. In the Right bound on # of Days To Shift field, enter the number of days after the original date within which the replacement datetime value must occur. For example, if you enter

To truncate the date, from the Generator Type dropdown list, select Date Truncation Generator.

With this option, in the output:

• The month and day are always January 1.

• If the original year is less than 90 years ago, then the year remains as is in the output.

• If the original year is 90 years or more ago, then the year is set to the current year minus 89.

Textual can parse datetime values that use either a format that Textual supports by default or a format that you add.

The Scramble Unrecognized Dates checkbox indicates how Textual should handle datetime values that it does not recognize.

By default, the checkbox is checked, and Textual scrambles those values.

To instead pass through the values without changing them, uncheck Scramble Unrecognized Dates.

By default, Textual is able to recognize a specific list of .

Under Additional Date Formats, you can add other datetime formats that you know are present in your data.

The formats must use a .

To add a format, type the format in the field, then click +.

To remove a format, click its delete icon.

In .docx and .xslx files, as long as the URL entity type handling option is not set to Off, Textual automatically changes the destination of hyperlinks to google.com.

On the Dataset settings page, the Word Document Settings section contains settings to determine how to manage .docx images, tables, and comments.

To display the Dataset settings page, on the dataset details page, click Project settings.

You can also use the Textual Agent to configure these options.

Configuring how to handle .docx images

For .docx images, including .svg files, you can configure the dataset to either:

• Redact the image content. When you select this option, Textual looks for and blocks out sensitive values in the image.

• Ignore the image.

• Replace the images with black boxes.

On the Dataset settings page, under Image settings for DOCX files:

• To redact the image content, click Redact contents of images using OCR. This is the default selection.

• To ignore the images entirely, click Ignore images during scan.

• To replace the images with black boxes, click Replace images from the output file with black boxes.

For .docx tables, you can configure the dataset to either:

• Redact the table content. When you select this option, Textual detects sensitive values and replaces them based on the entity type configuration.

• Block out all of the table cells. When you select this option, Textual places a black box over each table cell.

On the Dataset settings page, under Table settings for DOCX files:

• To redact the table content, click Redact content using the entity type configuration. This is the default selection.

• To block out the table content, click Block out all table cell content.

For comments in a .docx file, you can configure the dataset to either:

• Remove the comments from the file.

• Ignore the comments and leave them in the file.

On the Dataset settings page, to remove the comments, toggle Remove comments from the output file to the on position. This is the default configuration.

To ignore the comments, toggle Remove comments from the output file to the off position.

You can configure how Textual works with PDFs. For an individual PDF file, you can add manual overrides to selected areas of a file. Manual overrides can ignore detected values from Tonic Textual, or add redactions.

The PDF Settings section of the Dataset settings page provides options to configure how to work with PDF files.

You can also configure the signature, synthesis, and LLM classification mode options from the Textual Agent.

Configuring whether to redact PDF signatures

By default, Textual redacts scanned-in signatures in PDF files. You can configure the dataset to instead ignore the signatures.

Under PDF Settings:

• To redact PDF signatures, toggle Detect and redact signatures in PDFs to the on position. This is the default configuration.

• To ignore PDF signatures, toggle Detect and redact signatures in PDFs to the off position.

Textual has developed an updated synthesis process that is currently implemented for the following entity types:

• URLs

• Names

• Custom entity types

In particular, the new synthesis process improves the display of the synthesized values in PDF files. The values better match the available space and the original font.

Under PDF Settings, the New PDF synthesis mode (experimental) determines which process to use.

To use the new process, toggle the setting to the on position.

To improve the detection accuracy for sensitive values, you can optionally use an LLM to analyze structured components such as tables and form fields.

To enable this, under PDF Settings, toggle Use LLM to classify structured data for PII detection to the on position.

For PDFs as well as images, if multiple optical character recognition (OCR) models are available, you can select the specific model to use in the dataset. For information on how to enable specific models, go to .

Under PDF Settings, from the OCR Engine dropdown list, select the model to use.

From the Dataset settings page, you can import the following settings from another dataset:

• Project settings - The tags, .docx settings, and PDF settings.

• Entity settings - The entity type handling configuration and enabled custom entity types.

To import the settings:

1. On the dataset details, click Project settings.

2. On the Dataset settings page, click Import Dataset Settings.

1. On the Import Dataset Settings panel, from the Dataset dropdown list, select the dataset to import the settings from. To import settings from a dataset, you must have the View dataset settings permission for that dataset.

1. Under Settings to import:

   1. To import both the project and entity settings, click All settings.

   2. To only import the project settings, click Project settings only.

Tonic Textual tracks each change to the dataset configuration.

You can view the list of changes to the dataset configuration, and revert the dataset configuration to a selected version.

Displaying the dataset version history

To display the version history, click the clock icon at the top right of the dataset details page.

The Version history panel lists the versions in descending order, with the current version at the top of the list.

Displaying details for a version

For each version, the Version history list initially displays:

• A summary description of the change.

• The timestamp when the change occurred.

• The name of the user who made the change.

To display additional details about the change, expand the version description. For example, for a change to the handling option for an entity type, the details include the entity type, the previous handling option, and the new handling option.

You can restore an earlier version of the dataset. When you restore a version, Textual reverts the changes that occurred since that version.

On the expanded description of a version, to restore that version, click the Restore to <version number> button.

Textual adds a version to the history. The new version is marked as Restored, and the description indicates the version that you restored.

The expanded details for the version list the changes that Textual reverted when you restored the earlier version.

For a dataset that generates output files of the same type as the original file, the file preview displays the source and output versions of the file, and highlights the detected entity values.

From a file preview, you can change the entity type handling. For some file types, you can also add manual redactions.

De-identified view - output

For a de-identified file, the de-identified view reflects the output version of the file.

Entity values are redacted, synthesized, or ignored based on the dataset configuration.

In PDF and image files, values for entity types that use the Redact handling option are covered by black boxes.

In text files, redacted values are replaced with the entity type followed by a unique identifier. The values are highlighted to indicate whether the corresponding entity type redacted, synthesized, or ignored.

Original view - source

To display the original version of the file, click the original version icon. In original view, the detected entities are highlighted.

PDF and image files use a single color for the highlighting.

Text files use different highlighting colors to indicate whether the corresponding entity type is redacted, synthesized, or ignored.

To display the original and de-identified versions next to each other, click the side-by-side icon.

From a text, PDF, and image file preview, you can add manual redactions. When you add a manual redaction, you select the entity type to assign to it.

Adding manual redactions

Text files

To add a manual redaction to a text file:

1. In original view, select the text to redact.

2. On the redaction panel, from the dropdown list, select the entity type for the redaction.

PDF and image files

To add a manual redaction to a PDF or image file:

1. In original or side-by-side view, to change to redaction mode, click Add Redaction.

1. While in redaction mode, to add a redaction:

   1. Draw a box around the content to redact.

   2. On the redaction details panel, from the dropdown list, select the entity type for the redaction.

1. To exit redaction mode, click Done.

Textual displays the count of manual redactions on the current page.

To remove a manual redaction from a text file:

1. In either de-identified or original view, click the redaction.

2. Click Delete.

To remove a manual redaction from a PDF or image file:

1. In either de-identified or original view, click the redaction.

2. Click Delete Redaction.

Available column content types

When Textual processes a CSV file, it identifies the content of each column as one of the following:

• Sensitive - The column values are of a single type and format, and the values match an entity type.

• Not sensitive - The column values are of a single type and format, but the values do not match an entity type.

• Unstructured - Each column value might contain multiple entities of different types. For example, a column that contains notes or a longer description.

From the file preview, you can review and change the content type assignments for each column.

To change the assigned CSV column content type:

1. In the output table heading, click the dropdown icon.

1. On the configuration panel, click the column content type.

2. If you clicked Sensitive, then from the Column entity type dropdown list, select the entity type for the column values. Textual displays the current configured handling type for the selected entity type.

3. If you clicked Sensitive or Not Sensitive, then to save the changes, click Save. Changing the content type to Unstructured requires a rescan of the file. If you clicked Unstructured, then:

For each file in a dataset, you can download the output file.

Downloading a single output file

From the file list, to download a single output file, click the options menu for the file, then select the download option.

The dataset file preview also provides options to download the file.

Datasets that generate redacted output files

For datasets that generate redacted versions of the source files, in the file options menu, the option is Download File.

On the file preview, click the download icon.

Datasets that generate JSON output

For datasets that generate JSON output, you can download either the JSON output or the redacted version of the Markdown content.

For RTF files, you can also download the redacted content in HTML format.

To download all of the output files, click the download icon that is next to the file filter field.

For datasets that generate redacted versions of the source files, the download happens immediately.

For datasets that generate JSON output:

• To download the JSON output for the files, select Download JSON.

• To download the redacted Markdown content for all of the files, select Download Markdown.

• If the dataset contains any RTF files, then to download only the redacted HTML content for all of the RTF files in the dataset, select Download HTML (RTF only). If you select this option, then the download does not include other types of files.

Before you create guided redaction projects, configure the following options.

Expanding and collapsing the Textual Agent

The Textual Agent displays at the right side of the application.

When the Textual Agent is collapsed, to expand it, click the agent icon.

When the Textual Agent is expanded, to collapse it, click the agent icon.

While a regex-based entity type identifies values that match regular expressions, for a model-based custom entity type, you train a model to identify the entity values.

A model-based entity type is useful when the values are identified more by context than by format. For example, for an entity type that identifies the names of health conditions, it would not be possible to set up regular expressions that identify the values.

You iterate over text-based guidelines that identify the entity type values in a smaller set of data, then use a larger set of data to train one or more models.

Each trained model is based on a selected version of the guidelines.

You select the trained model to use for the custom entity type.

For a custom model entity type, the overall process is as follows:

The first step is to identify the entity values that are in a small set of test files. The test files and established values are used both to iterate over the model guidelines and to assess how well your trained models perform.

1. When you create the model-based custom entity type, you . For example, "Scientific names of health conditions". The description is the first version of the model guidelines. The guidelines tell the model how to identify the entity type values.

Textual provides access to a library of entity types that we created from trained models.

These entity types cover a range of additional types of values. You might be able to use these types instead of adding custom types.

You can search for and select types to import that apply to your datasets. The imported types display in the list of custom entity types.

You then .

To start an import, on the Entity Types page, either:

• Click Import type

For datasets that produce redacted files, for each entity type, you choose how to handle the detected values. This determines how each value displays in the output files.

For datasets that create JSON output, the entity type handling determines the display in downloaded Markdown or HTML files.

In addition to the dataset details page options, you can also use the to configure the entity type handling.

The available options are:

• Synthesize - Indicates to replace the value with another realistic value. For example, the first name value Michael might be replaced with the value John. The synthesized values are always consistent, meaning that a given entity value always has the same replacement value. For example, if the first name Michael appears multiple times in the text, it is always replaced with John. Textual does not synthesize any excluded values. For custom entity types, Textual scrambles the values.

For PDF files, you can add manual overrides to the initial detections, which are based on the detected data types and handling configuration.

For each manual override, you select an area of the file.

For the selected area, you can either:

• Ignore any automatically detected values. For example, a scanned form might show an example or boilerplate content that doesn't actually contain sensitive values.

The Textual guided redaction tool blocks out sensitive values in files. For example, you might use guided redaction to prepare documents to provide in response to a Freedom of Information Act request.

Guided redaction supports built-in and custom entity types. Textual completes an initial scan of the files to identify sensitive values in the files. You can then manually add and remove redactions.

The values are covered with a black or white box. You assign and display reference codes to identify the type of content for each redaction.

The overall workflow is as follows:

Before you create a guided redaction project:

The Guided Redaction page contains the list of redaction projects.

To display the Guided Redaction page, in the Textual navigation menu, click the guided redaction icon.

The list displays the projects that you have access to.

For each project, the list includes:

• The name of the project

Reference codes are used to indicate the type of value that is redacted.

Before you can start a guided redaction, you must set up the reference codes.

You can optionally link each reference code to one or more Textual entity types. Make sure that you map reference codes to all of the entity types that appear in your files.

On the Guided Redaction page, to display the list of reference codes, click Reference Codes Settings.

For each each reference code, the list includes:

Before you start training models, on the Model data setup page, you select the training data to use.

About training data

The training data is a much larger set of files than the test data, and can include hundreds of files or more. The data should ideally contain at least 1,000 values for the entity type. For example, for an entity type to identify health conditions, you might use 5 medical appointment reports in your test data, but several hundred medical reports for your training data.

Similar to the test files, the training files should be relatively small - no more than 5,000 words.

For training data, there is no option to paste in text. Training data files are either uploaded from a local file system or selected from a cloud storage solution.

If you selected the test data from a cloud storage solution, then you must use the same cloud storage solution for the training data. For example, if you selected the test data from Amazon S3, then you must select the training data from Amazon S3.

You can add files to the training data at any time. New files are only used for models that are trained after the files are added.

If there are no training files, then on the Model data setup page, click Upload files, then search for and select the files to upload.

To add more uploaded files to the training data:

1. Click Add training data.

1. Click Upload files.

2. Search for and select the files to add to the training data.

If there are no training files, then on the Model data setup page, click the cloud storage solution to use, then select the files to add.

If the test data came from a cloud storage solution, then you must use the same cloud storage option for the training data.

To add more cloud storage files to the training data:

1. Click Add training data.

1. Select the cloud storage solution.

2. Select the files to add to the training data.

For training data, you can select entire folders. Textual then adds all of the files in the folder.

On the Model data setup page, to display the content of an uploaded file, click the file name.

Each training data file goes through the following statuses:

• Queued for upload - The file is not yet uploaded.

• Uploading - Textual is uploading the file.

• Ready - The file is uploaded and is used for subsequent model training.

Model training cannot start until all of the currently uploaded files are Ready.

On the Model data setup page, to delete a training file:

1. Click its delete icon.

2. On the confirmation panel, you can choose to skip the confirmation when you delete training files. If you select this option, then the next time you delete a training file, the file is deleted immediately, and the panel does not display.

3. Click Delete.

When you delete a training file:

• For existing models that annotated the file:

  • The entity counts continue to reflect the entities that were detected in the file

  • The file name remains in the list on the model details.

For a cloud storage dataset, you can:

• Update the cloud storage credentials. Note that this option is only available if you provided the credentials manually. If you use the credentials set in environment variables, then you cannot change the credentials.

• Change the output location for the generated output files.

You configure the connection credentials and output location from the Dataset settings page. To display the Dataset settings page, on the dataset details page, click Project settings.

After you update the configuration, click Save Dataset.

From the credentials section, to update the cloud storage credentials, click Update <Cloud storage solution> Credentials.

On a self-hosted instance, you cannot change the credentials location. For example, you cannot change the location from User to Environment or from Environment to Assume Role.

To provide updated user credentials for Amazon S3:

1. In the Access Key field, provide an AWS access key that is associated with an IAM user or role. For an example of a role that has the required permissions for an Amazon S3 dataset, go to .

2. In the Access Secret field, provide the secret key that is associated with the access key.

3. From the Region dropdown list, select the AWS Region to send the authentication request to.

To update the assumed role configuration:

1. In the Role ARN field, provide the Amazon Resource Name (ARN) for the role.

2. Under External ID Uniqueness, select how to determine the uniqueness of the external ID. The ID can be unique for the organization or for the dataset.

3. In the Session Name field, provide the role session name. If the session name is empty, then Textual generates the session name.

To provide updated credentials for Azure:

1. In the Account Name field, provide the name of your Azure account.

2. In the Account Key field, provide the access key for your Azure account.

3. To test the connection, click Test Azure Connection.

SharePoint credentials must have the following application permissions (not delegated permissions):

• Files.Read.All - To see the SharePoint files

• Files.ReadWrite.All -To write redacted files and metadata back to SharePoint

• Sites.ReadWrite.All - To view and modify the SharePoint sites

To provide updated credentials for SharePoint:

1. In the Tenant ID field, provide the SharePoint tenant identifier for the SharePoint site.

2. In the Client ID field, provide the client identifier for the SharePoint site.

3. In the Client Secret field, provide the secret to use to connect to the SharePoint site.

The output location is where Textual writes the redacted files.

When you create a cloud storage database, after you select the initial set of files and folders, Textual prompts you to select the output location.

For an existing dataset, you set the output location from the Output Location section of the Dataset settings page.

Click the edit icon, then select the cloud storage folder where Textual writes the output files for the dataset.

When you generate output for a cloud storage dataset, Textual creates a folder in the output location. The folder name is the identifier of the job that generated the files.

Within the job folder, Textual recreates the folder structure for the original files.

Textual then writes the output files to the corresponding folders.

For a cloud storage dataset, you manage files from the file selection panel.

When you create a dataset, after you provide the cloud storage credentials and save the dataset, Textual immediately prompts you to select dataset files. After you select the files and click Next, Textual prompts you to set the output location.

For an existing dataset, to display the File Selection panel

1. On the dataset details page, click Project files.

2. On the dataset files page, click Select Files.

The file selection includes:

• Whether to restrict the dataset to specific file types

• The files or folders to include in the dataset

When you change the file selection, Textual scans the files for entities. For more information, go to .

When you select files, you can filter the selectable files based on file extension.

To limit the file extensions to include:

1. Click File Extension Filter. By default, all file extensions are included, and none of the checkboxes are checked.

2. Check the checkbox for each file extension to include. As you select the file extensions to include, Textual updates the navigation pane so that you can only select files that have one of those file extensions. It hides files that have other file extensions and folders that do not contain files with the selected file extensions.

In the file selection area, you navigate to and select the folders and files to add to the dataset.

In the navigation area, to display the contents of a folder, click the Open link for the folder.

To add a folder or file to the dataset, check its checkbox.

In the navigation pane, when you check a folder checkbox, Textual adds it to the Prefix Patterns list.

Instead of navigating to a folder and selecting it, you can add the path to the list manually

To add a folder path:

1. Click Add Prefix Pattern.

2. In the field, type the path to the folder, then click the save icon.

To remove a folder path from the dataset, either:

• In the navigation pane, uncheck its checkbox.

• In the Prefix Patterns list, click its delete icon.

For the selected folders, the dataset includes all of the applicable files in the folder that:

• Are of a file type that Textual supports

• Match the file extension filter

In the navigation pane, when you select an individual file, Textual adds it to the Selected Files list.

To delete a file, either:

• In the navigation pane, uncheck its checkbox.

• In the Selected Files list, click its delete icon.

As you redact and review the project files, you set the status of each file and project. Projects and files use the same set of status values.

Each status is associated with a color.

Built-in statuses

Textual comes with a set of built-in statuses. The built-in statuses are:

• Not started - This is the default status. It is applied automatically to all new projects and files. You cannot delete this status.

• In progress

• Ready for review

• Review in progress

• Done - This is intended to be the final status for a project or file. You cannot delete this status.

You can change the name and assigned color of all of the built-in statuses. You can delete the built-in statuses that you do not need, except for the Not started status and the Done status.

You can also add custom statuses, to accommodate your particular redaction and review process.

On the Guided Redaction page, to display the status list, click Status settings.

The status list shows the status name and the associated color.

From the Status Settings panel, to add a status:

1. Click Add a status.

2. On the status configuration panel, in the field, provide a name for the status. Status names must be unique.

3. Click the color to assign to the status.

4. Click

From the Status Settings panel, to change the status configuration:

1. Click the status name.

2. On the status configuration panel, you can change the status name and color. Remember that the status name must be unique.

1. To save the changes, click Save.

You cannot delete:

• The built-in Not started status.

• The built-in Done status.

• A status that is currently assigned to a project or file.

From the Status Settings panel, to delete a status:

1. Click the status name.

2. On the status configuration panel, click the delete icon.

The audit log provides a history of actions associated with a dataset.

Displaying the audit log

To display the audit log for a dataset, on the dataset details page, click Audit log.

Each entry in the audit log includes:

• The timestamp when the event occurred.

• The user who performed the action.

• How the action was performed. For example, a change might be made directly in the application, using the SDK, or using the MCP server.

• The type of action.

• Additional details for the action.

The audit log includes the following types of actions:

• Dataset created

• Dataset deleted

• Dataset settings changed

• Dataset settings imported

To download the audit log to a .csv file, on the Audit log page, click Download CSV.

On the Guidelines refinement page, you prepare the guidelines that define a model.

To test each version of the guidelines, Textual uses the guidelines to detect entity values in the test data. It then generates scores to indicate how closely those detection results match the values that you established during your initial review.

Textual also generates recommendations to improve the guidelines.

To work on the guidelines, click Guidelines refinement. The Guidelines refinement option is enabled when you complete the review on the initial set of test files.

The first time you display the Guidelines refinement page, Textual uses the guidelines that you provided during the entity type creation to populate the Version 1 tab.

After you select your training data, on the Model training page, you create one or more trained models.

For each model, you select the version of the guidelines to use. Textual first uses those guidelines to annotate the training data. Based on how well the guidelines identified the values in the training data, you decide whether to start the model training.

When the training is complete, the model scans the test data. The model is scored based on how well it detected the definitive values that you confirmed in the test data.

For each model, the model list includes:

• Model

In a dataset, for each built-in entity type, you can configure additional values to detect, and values to exclude. You cannot define added and excluded values for custom entity types.

You might add values that Textual does not detect because, for example, they are specific to your organization or industry.

You might exclude a value because:

• Textual labeled the value incorrectly.

• You do not want to redact a specific value. For example, you might want to preserve known test values.

The entity types list includes the following additional entity types that you activated for the dataset:

• Custom regex-based and model-based types.

• Types that you .

From the Entity settings page, you can enable and disable additional types.

You can also update the configuration of a regex-based type, and go to the details page for a model-based type.

By default, when you select the Synthesize option for a custom entity type, Textual scrambles the original value.

From the generator dropdown list, select the generator to use to create the replacement value.

The available generators are:

A dataset might contain multiple files that have the same structure, such as a set of scanned-in forms.

Instead of adding the same manual overrides for each file, you can use a PDF file in the dataset to create a template that you can apply to other PDF files in the dataset.

When you , you can apply a template.

To add a PDF template to a dataset:

1. On the Dataset settings page, under PDF Settings, click PDF Templates.