Education

Step 5 - Add Semantic Relationships to the Definition

Once you are finished with your definition you’ll need to place the new term into context with other terms. It allows your readers to see how terms interact with each other. It allows Natural Language Processing Engines to relate terms together. It is the core in pattern-matching for harmonizing regulatory structures to each other.

Basic semantic relationships

The following basic relationships have been taken from the Simple Knowledge Organization System’s (SKOS) Mapping Vocabulary Specification[1], as shown below.

Basic semantic relationships

View Larger

They offer the ability to distinguish subtle relationships between two terms. As stated in the specification, “Many knowledge organization systems, such as thesauri, taxonomies, classification schemes and subject heading systems, share a similar structure, and are used in similar applications. SKOS captures much of this similarity and makes it explicit, to enable data and technology sharing across diverse applications.”

has-exact-match

If two concepts are an exact match, then the set of resources properly indexed against the first concept is identical to the set of resources properly indexed against the second. Therefore, the two concepts may be interchanged in queries and subject-based indexes. (Is inverse with itself.)

has-broad-match

If “concept A has-broad-match concept B,” then the set of resources properly indexed against concept A is a subset of the set of resources properly indexed against concept B. (Is inverse of has-narrow-match.)

has-narrow-match

If “concept A has-narrow-match concept B,” then the set of resources properly indexed against concept A is a superset of the set of resources properly indexed against concept B. (Is inverse of has-broad-match.)

has-major-match

If “concept A has-major-match concept B,” then the set of resources properly indexed against concept A shares more than 50% of its members with the set of resources properly indexed against concept B. (No inverse relation can be inferred.)

has-minor-match

If “concept A has-minor-match concept B,” then the set of resources properly indexed against concept A shares less than 50% but greater than 0 of its members with the set of resources properly indexed against concept B. (No inverse relation can be inferred.)

The limitations with basic semantic relationships based off the SKOS model

The problem in the SKOS model is relationships are limited to a single term or a single phrase. This model is great if you want to know that draft or chart is the same as map or not as broad as interpret. Basically, you are limited to three categories for practical purposes; broader, same, and narrower as shown in the diagram below.

Visual SKOS Model

View Larger

What the SKOS and basic semantic relationship model doesn’t tell you is why interpret is a broader concept, or why scale is a narrower concept. What they don’t show are the linguistic relationships between the terms.

To extend the relationships past broader, same, and narrower, you’ll need a more advanced semantic relationship system. It should consider real world relationships such as one concept being a category for another concept, or one concept enforcing another concept, or even one concept including another concept as a part of it (versus the parent being a category). The illustration that follows re-examines the semantic relationships of the term map, shown above, using a more advanced set of semantic relationships. These relationships provide a much more robust understanding of connecting terms than a simple broader, same as, and narrower model can provide. Advanced semantic relationships extend the model by adding linguistic and conceptual connections to each relationship.

Advanced semantic relationships

View Larger

Advanced Semantic Mappings

There are many more relationships you’ll need to put into place if you want to provide greater context for your readers or Natural Language Processing Engine. Here are a few more of the relationships you’ll need.

Synonyms and Antonyms instead of has-exact-match

Synonyms are broader than exact matches, as they extend the relationship to facts or states of having correlation, interrelation, materiality, conformity, and pertinence between concept A and concept B. And antonyms then have enough variability, incongruence, and disassociation to be their opposite. The antonym is the inverse of the synonym and vice versa.

Metonymy

Included in the type of synonyms is metonymy, the semantic relationship that exists between two words (or a word and an expression) in which one of the words is metaphorically used in place of the other word (or expression) in particular contexts to convey the same meaning.

Included in the category of antonyms are complementary pairs, gradable pairs, and relational opposites[2].

Complementary pairs

Complementary pairs are antonyms in which the presence of one quality or state signifies the absence of the other and vice versa. A couple of samples are single/ married, not pregnant/pregnant. There are no intermediate states in complementary pairs.

Gradable pairs

Gradable pairs are antonyms which allow for a natural, gradual transition between two poles. A couple of examples are good/bad, hot/cold. It is possible to be a little cold or very cold, etc.

Relational opposites

Relational opposites are antonyms which share the same semantic features, only the focus, or direction, is reversed. A couple of examples are tied/untied, buy/sell, give/receive, teacher/pupil, father/son, and open/refrain from opening.

Non-standard forms of has-exact-match

A spigot and a faucet are two defined words that are exact matches, or synonyms, of each other. That’s an easy rule to implement. However, language is messy, and the uses of language within compliance documents is even messier. That’s why you must have advanced rules that go beyond synonyms for use cases such as a personal data request being called a request for personal data, an information request from the data controller, or even a request for information on the processing of personal data. To handle these types of use cases you must have a semantic rule that says “if the definition of a term-of-art matches the definition of a previously accepted dictionary term, the term-of-art should be considered an exact match and therefore be labelled a non-standard representation of the accepted term”.

Replacing the broad and narrow matches with more specific categorization

The major and minor relationships described in the SKOS model are limited to linguistic parents and their children (or half children as a minor match might be thought of). However, there are many relationships that are more specific that can and should be applied, especially when working with named entities and leveraging a Natural Language Processor’s named entity recognition engine. By replacing the simple broader and narrower matches with more specific categorization, you can achieve structures like those employed by the Compliance Dictionary, as shown below.

RelationshipDescriptionExamples
Linguistic ParentTerms that are linguistically broader than the focus term, including origins of terms.TermSenior Systems Analyst
Linguistic Parent – systems analyst, senior
Linguistic ChildTerms that are linguistically narrower than the focus term, including derivatives. This is the in-verse of Linguistic Parent.Term - systems analyst
Linguistic Child - Senior Systems Analyst
Category ForA term of which the focus term is a kind of.Term – tablet
Category For – portable electronic device
Type ofTerms that are kinds or examples of the focus term. This is the inverse of Category For.Term – portable electronic device
Type of – laptop
IncludesTerms the focus term is an element of. It is the same as hyponymy.Term – Personally Identifiable In-formation
Includes – mailing address, individual’s Social Security Number
Part ofTerms whose definitions are an element of the focus term. This is the inverse of IncludesTerm – Personally Identifiable Information
Part of – privacy related information
Used to CreateA term that is a template for or used to create the focus term.Term – UCF Mapper software
Used to Create – Authority Document mapping
Is Created byA term that is comes from or is generated by the focus term. This is the inverse of Used to Create.Term – system audit report
Is Created by – Secure Configuration Management Tool
Is Referenced byA term that mentions or references the focus term.Term – evidence           
Is Referenced by – probable cause
ReferencesA term that the focus term mentions or cites. This is the inverse of Is Referenced by.Term – evidence
References – business exception rule
Used to EnforceA term that uses the focus term to happen or cause compliance.Term – configuration rule
Used to Enforcesystem configuration
Is Enforced byA term that uses the focus term to happen or cause compliance. This is the inverse of Used to Enforce.Term – PCI-DSS
Is Enforced by – payment brand
Used to PreventA term that prevents the focus term.Term – sanctions
Used to Prevent -– unauthorized data processing
Is Prevented byA term that is prevented by the focus term. This is the inverse of Used to PreventTerm – stealing
Is Prevented byarmed guard

Questions for analyzing the relationships of your terms

As of this writing, there isn’t a computer system that will automatically analyze terms, even in their context within a document, and determine what the relationships should be. At best, they are running between 40-45% accurate[3]. This means you’ll want to manually ask yourself the questions, which isn’t really that hard[4]. Here’s our cheat sheet for you.

Relation TypeQuestions
SynonymsHave you seen this term spelled differently?
Have you seen this term written completely different (Personally Identifiable Information/individual’s non-public data)?
Is this a metaphor for another term?
Are there metaphors for this term?
AntonymsAre there any qualities of this term that signify the absence of qualities of another term (single/married)?
Could this term be graded on a spectrum (hot/cold)?
Is there an opposite relationship of this term (tied/untied)?
Category ofWhat terms fall under this category?
Type ofAre there any other examples of this term?
IncludesWhat does this term include?
Part ofIs this term a part of a greater whole?
ReferencesDoes this term refer to other terms?
Is this term referenced by other terms?

By creating semantic relationships to your definitions, the reader will be able to understand how the term works with other terms.

Continue Reading


[1] https://www.w3.org/2004/02/skos/mapping/spec/ and https://www.w3.org/TR/skos-reference/

[2] “Linguistics 201: Study Sheet for Semantics.”

[3] Malaise, Zweigenbaum, and Bachimont, “Detecting Semantic Relations between Terms in Definitions.”

[4] Storey, “Understanding Semantic Relationships.”

Get the Definitions Book

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.