5 Ways Semantic Models Help Protect Sensitive Data

5 Ways Semantic Models Help Protect Sensitive Data

Posted on March 29, 2024 0 Comments
decorative image

Progress Semaphore clients include some of the largest companies in the world across many different industries, and all of them—big or small—create and process extraordinary amounts of data every day. One of their most important challenges is protecting data considered sensitive.

Sensitive data is any information that needs safeguarding due to the potential negative consequences on an organization's well-being, privacy, reputation or security if it is lost, misused, altered or accessed without authorization.

All organizations grapple with the challenge of identifying and protecting sensitive data within the vast expanse of their data repositories. Identifying this data in the unstructured data in documents that comprise 80% of an organization’s information is especially difficult. Semantic models have emerged as powerful tools for organizing and representing data in a meaningful form, which can be extended to specific use cases like identifying and helping secure sensitive information.

Traditional Approaches to Managing Sensitive Data

All organizations have the challenge of protecting their sensitive information. But in looking to address this challenge, they face a trade-off. They often need to provide access to documents to those who need it for their business function, but they also must be able to protect information from unauthorized access—whether external or internal. Suppose they take a blanket approach and try to protect everything to the highest levels of security. In that case, they will make valuable data unavailable or difficult to access for the users who need it.

The answer to this problem is to classify data in alignment with your organization’s security regime. This allows you to govern who has access to what data and only apply the expense of enhanced data protection to the data that requires it.

When a company considers Progress Semaphore for help with this challenge, they are generally using a few traditional approaches. One is to have users leverage a software tool to manually “tag” documents with the desired classification. There are problems with this approach in that it is time-consuming and creates inconsistent and often incorrect classification results. A second approach is to search for regular expressions (REGEX) that look for specific matches to a predefined list of terms. This approach has the benefit of being an automated way to classify data, but it also may miss many of the proper classifications as it will only work for an exact match for something on the list.

A third strategy is to use a pure machine-learning approach. While initially, this appears to require less effort, in actuality, it requires careful selection and pre-labeling of massive training sets that represent every potential category of protection, plus extensive review processes of the results. In the end, the machine's decisions are opaque and unexplainable—which is not helpful when faced with challenges from regulators and lawyers.

Understanding Semantic Data Models

Semantic models, built on the principles of natural language understanding, delve beyond the surface of words and help you capture meaning from your data in a single enterprise knowledge graph that can be used for the entire organization. They can be used to comprehend context, relationships and concepts unique to your organization, making them invaluable in pinpointing sensitive data. With semantic data modeling, companies can reorganize their data by defining the real-world entities within it and their relationships from a business user’s perspective.

Components of Semantic Data Models

Concepts and Entities

Semantic data models consist of concepts and entities that represent real-world objects, events or ideas and serve as the building blocks of the model.

Attributes and Properties

Semantic data models have attributes and properties that provide additional details and context about the data, facilitating a profound understanding of the represented data.


Semantic data models help organize data by capturing relationships between entities and connections within the given context.

Semantic models make data more valuable by creating a common language that both computer systems and people can understand.

Applying Semantic Data Models for Data Identification

With semantic data models, businesses can decipher information more successfully, leading to better decisions based on more complete and contextualized data. They can help businesses with:

  1. Automated Document Classification - Semantic models contribute to automated document classification. By understanding the context and content of documents, these models can classify them based on predefined categories—a fundamental step in identifying sensitive data.
  2. Metadata Enrichment - Organizations can create a structured system that enables easy classification and identification by tagging sensitive data with appropriate metadata. They extract key information from documents, turning raw data into a structured narrative. This metadata is used to identify sensitive content. Try metadata management
  3. Named Entity Recognition (NER) - Named Entity Recognition allows these models to identify entities like names, locations, dates, etc. NER becomes a powerful tool to spot personally identifiable information (PII) and other critical entities.
  4. Relationship Recognition - It's not just about individual entities; it's about understanding relationships. Semantic models can decipher connections between entities, unveiling patterns that might indicate sensitive information. For instance, a person’s name appearing in a resume vs. an employee record can have very different sensitivity implications.

Benefits of Semantic Models in Sensitive Data Identification

Semantic data models are gaining popularity in different industries due to their ability to establish and enforce data governance policies more effectively. They can provide several benefits for sensitive data identification, particularly in the context of data security, including:

  1. Contextual Data Understanding – By using the meaning and context of data, semantic models make it easy for businesses to understand and interpret information. They don't merely see words; they interpret the context in which those words exist.
  2. Transparency in Classification – Semantic models enhance accuracy and transparency when classifying documents. They allow for precise categorization that, unlike pure machine-learning approaches, is traced back to a knowledge model that clearly reflects the business strategies and policies that inform it.
  3. Real-time Analysis – As data evolves and business priorities change, these models adapt so that identifying sensitive information is not a static process but a dynamic, ever-evolving one.
  4. Better reporting – Semantic models don't just look for explicit markers of sensitivity; they can also classify data by important business concepts of product, process, project and topic, resulting in a complete view of an organization’s data and documents.
  5. Security – By identifying sensitive information in documents, data cybersecurity and records management teams can more easily process and protect them according to organizational security requirements.

Semantic knowledge models can be particularly useful when working with large data sets where quickly and accurately identifying sensitive information is essential for better compliance and security.

Moving Towards a Semantic Future in Data Governance

In the critical pursuit of securing sensitive information, semantic models emerge as powerful tools in the data management and governance space. These tools have clear advantages that can supplement or replace other approaches like pattern recognition, manual tagging and machine learning. Their ability to understand context, transparently classify and adapt to changing organizational mandates or regulatory regimes, positions them as crucial components in the complex challenge of sensitive data identification. As organizations navigate the evolving world of data security, integrating semantic models becomes not just a choice but a strategic imperative. Let semantic models help you move toward a future where sensitive information is not just better protected—but truly understood.

Discover how Semaphore can help you identify sensitive data in your business context.

Doug Dunn

Doug Dunn is a Senior Enterprise Account Manager on the Progress Semaphore team. Doug helps global customers reveal smarter decisions by providing a unified, Semantic AI data platform that delivers comprehensive insights.

Outside the office, Doug is an avid runner and also enjoys the beauty and challenge of mountain backpacking.


Comments are disabled in preview mode.

Sitefinity Training and Certification Now Available.

Let our experts teach you how to use Sitefinity's best-in-class features to deliver compelling digital experiences.

Learn More
Latest Stories
in Your Inbox

Subscribe to get all the news, info and tutorials you need to build better business apps and sites

Loading animation