Relationship Extraction is a Critical AI Technology for Effective Link Analysis

What is Link Analysis?

Link Analysis is a well-known technology for knowledge discovery. It discovers relationships between entities in data using network visualizations.

Typical relationships that are of interest in many domains include those between:

• Person and organization (e.g., a person is employed by an organization)
• Person and person (e.g., a person is someone’s brother)
• Organization and organization (e.g., an organization is a subsidiary of another)
• Person and place (e.g., a person lives in a certain town)
• Etc.

Link Analysis might use a “six degrees of separation” style of search where the search can be extended out through multiple nodes in the network.

Link Analysis is Critically Important in Many Domains

Since Link Analysis has proven helpful to humans in discovering significant and otherwise hidden links in data, it is used extensively in many different domains such as the following:

• Law Enforcement
• National Security
• Due Diligence for Risk Management
• Fraud Detection
• Business intelligence

Automated Relationship Extraction is Now Critical to Support Link Analysis

Given the stupendous and ever-growing amount of unstructured data now available on the Internet in the form of news, corporate reports, social media, traditional blogs, etc., the traditional method of performing link analysis is now no longer viable. In the past, analysts would read documents and manually build out link charts of relationships between entities. This approach does not scale to the present overwhelming amount of unstructured data and is a considerable waste of analysts’ valuable time.

What is needed is a technology that can analyze unstructured data and build charts of relationships automatically. There is such a technology: Relationship Extraction.

How Relationship Extraction Facilitates Link Analysis

Relationship Extraction automates the identification of critical entities and links between them in unstructured text. It specifies the precise semantic relationship between entities based on a defined ontology of relationships. It outputs a structured representation of these relationships, which can be stored in a database. A link analysis tool can then import the structured data and visualize it for analysts.

How Relationship Extraction Works

Here is an example of how Relationship Extraction works:

• Council Dr. Bibek Debroy, the chairman of the Economic Advisory Counsel

Relationship Extraction uses a semantic ontology of predefined relationships. It labels each occurrence of a relationship according to the ontology, e.g., RELATIONSHIP TYPE: AFFILIATION for the example above. This label expresses the affiliation that Dr. Bibek Debroy has with the Economic Advisory Counsel.

Relationship Extraction also identifies the role of each participant in the relationship with the appropriate label, e.g., Bibek Debroy as the EMPLOYEE and the Economic Advisory Council as the EMPLOYER in the example above.

Relationship Extraction handles the linguistic variation found in natural language, which can express the same concept in many different ways. For instance:

• The chairman of the Economic Advisory Council, Dr. Bibek Debroy.
• Bibek Debroy was appointed chairman of the Economic Advisory Council
• The president hired Bibek Debroy to head the Economic Advisory Council
• Debroy led the Economic Advisory Council

 Extraction Enhances the Process of Knowledge Discovery

Relationship Extraction supports effective and large-scale Knowledge Discovery. It enables the scenarios in different industries such as the following:

• A business analyst can study competitors by having Relationship Extraction process business news articles and press releases containing information about hirings and departures of executives.
• A government intelligence analyst can generate charts of the membership of terrorist groups by having Relationship Extraction process cable message traffic and other intelligence sources.
• A law enforcement analyst can map out the membership of criminal gangs based on Relationship Extraction having processed crime reports.

Summary

Relationship Extraction identifies crucial links between entities in a large amount of unstructured text. The combination of Link Analysis and Relationship Extraction greatly facilitates knowledge discovery and at the same time accelerates producing such knowledge.