Custom data discovery and enforcement using rule templates (data concepts and semantics)
A data concept is the class or family of a dataset for example: Loan Data, Stock Data, Position Limit Data, Retail Product Data, Patient Health Data, Interest Rate Data, etc... By giving data classes or "classifying" datasets we can transfer common understanding, rules and ML to datasets. This powerful technique allows a data steward to set up concepts once and enables the entire organization the ability to unify and standardize on common rules and terms, solving many metadata scale challenges.
Security Reference Data - Bloomberg financial data Home Loan Data - Mortgage application data
EMAIL, ZIP CODE, SSN, CUSIP, GENDER, ADDRESS, CURRENCY CD, SKU, EIN, IP ADDRESS, PHONE, LICENSE, VIN, CREDIT CARD
A semantic is the "semantic type" of a column or attribute of a dataset. All columns have a physical type such as String, Int, Date etc... but the semantic understanding of what type of String is in the column can be very important. It also allows us to enforce DQ validation rules out of the box.
Owl's semantic scanning self identifies standard columns and automatically provides the proper protection. This makes it easy to get started adding common rules for specific use-cases.
Owl offers out of the box rules for 1-click rule creation
With the Run Discovery modal, you can run a DQ Scan to detect for the semantics assigned to a selected data concept. The Run Discovery algorithm automatically selects the best match if a column matches two or more data classes. Data class match criteria are determined by percent match and name distance.
You can access the Run Discovery feature via:
In Catalog, select your dataset.
In the Actions dropdown menu, click Data Concept.
Select an option from the Data Concept dropdown and click Run Discovery.
Via DQ Job
From the DQ Job page, select your DQ Job.
Click the Rules tab in your DQ Job.
Click the Rule Discovery button.
In the Data Concept window, select your Data Concept.
Click Run Discovery.
PII - personally identifiable information MNPI - materially non public information
PCI - credit information like a credit card number PHI - HIPAA medical information
Data Discovery: The Power of Combining All Three into One Domain
Now imagine if you could classify your datasets as concepts, then automatically have all the columns be recognized semantically(with validation rules in place) as well as have the columns labeled with sensitivity tags. It might look something like the below.
Steps To Use
Step 1: Create DQ Job with Semantic On
To start, create a new DQ Job and select Semantic ON on the Profile options page
Step 2: In Catalog, select and apply your Data Concept
Navigate to your dataset in Catalog and select the Data Concept you would like to apply with the Actions dropdown menu.
See below sections on how to Administering Data Concepts as well as how to Create and Manage Semantics
Step 3: Rerun your DQ Job with applied Data Concept
Please rerun your DQ Job so that Collibra DQ can 1) profile your data, 2) auto-generate the rules based on the Semantics under the Data Concept, and 3) highlight any break records.
Success! Review Findings
On the Profile page, please observe the newly tagged Semantics on the applicable columns
On the DQ Job page, please browse your newly created rules based on Semantics as well as any corresponding rule breaks
Creating and Managing Semantics
Create, test, and manage your Semantics in Collibra DQ in your Rule Builder wizard on the Create Generic Rule tab. Below is an example of creating a RegEx Semantic
Administering Data Concepts
Setup your data concepts once and let the entire organization benefit by unifying all datasets to a common understanding in the admin Data Concepts page.
Physical Schemas to Semantics
Below you can see the benefit of organized metadata. PDEs or physical data elements organized/tagged by semantics. This allows for sub-second searches while in catalog or searching for data to figure out where all your PII data lives, or what systems have "loan data".
Above you can see Data Concepts in Yellow, Semantics in Gray and Sensitive labels in Orange. Enabling you to organize all your data in classes, search and discover types no matter what system they live in or what the PDE column name is. Transforming technical types into business metadata.
Business Unit Roll up Reporting
Now that we have all PDEs discovered and tagged and rolled up into business terms, we can roll up technical assets like database tables and files into business reports across departments and non technical concepts.