The hidden world of deep semantics

In our quest to build and grow the enterprise of the future, we should always keep an eye on what’s going on in the knowledge sciences. One of the more interesting places to explore in that mysterious realm is the nature of deep structure in human language and speech, called deep semantics.

The goal of semantic analysis is to decipher the intent or meaning behind a narrative. Aided by massive computational power, current brute-force approaches attempt to do that by analyzing millions of combinations of word stems and grammatical structures.

But as we’ve discussed in our recent article on big data, sometimes less is better. The good news is we can streamline our approach by focusing on only a few dozen motor reflex arcs called phonemes. They are the basic elements (atoms) of speech that our physiology makes available for us to convey an infinite number of possible words. Combine them with only a few dozen semantic primes (see Anna Wierzbicka’s book, Semantics: Primes and Universals, 1996), and we arrive at a more manageable set of conceptual structures.

A close analogy would be the 118 elements in the periodic table. The rules of valence define which atoms can be combined and which ones can’t. Similarly, by uncovering the rules of aggregation of the basic elements of human speech, we can move smartly toward “less is better” in our approach to semantic analysis.

From gobbledygook to muhkam

Realizing the significance of such fundamental structures, Tom Adi of the Readware Institute has spent decades attempting to untangle the semantic Gordian knot. A breakthrough occurred when he analyzed classical Arabic in portions of the Qur’an known as muhkam—verses rooted in logic and wisdom.

Adi found that when compared to other writings, muhkam verses exhibited the highest degree of semantic consistency. His analysis of such verses resulted in identifying a stable set of seven process types and four orientations, with each process/orientation pair linked directly to one of 28 Arabic phonemes (see “Muhkam Algorithmic Models of Real World Processes for Intelligent Technologies,” The International Journal of Robotics Applications and Technologies, July-December 2013). For example, the sound “qaf” indicates the process of containment, closed to others. “Ha” expresses assignment, open to others.

He also discovered a dozen rules for aggregation of those basic sounds into groups of three, the common structure for most Arabic words. Overall, he’s identified roughly 50 semantic primes, such as wishing for or wanting something, sadness (low and high intensity), prejudice and fear. As you might expect, the emotional flavors of a concept (love/hate, joy/sorrow, etc.) are expressed by various formations of the lips, teeth, hard and soft palates, and throat (smooth vs. percussive, open vs. closed, etc.).

Sanskrit, one of the oldest languages in the world, has similar characteristics. Its 25 major consonants, along with a few minor variations, comprise all possible combinations of muscular formations that can be used to produce speech—a periodic table of the elements for speech, if you will.

Interestingly, a Sanskrit word never represents an object. Rather, every word represents a set of properties. That is pure knowledge representation, where a word conveys the experience of observing an object, including its effects on the observer.

Let’s see how we can take these ideas and use them to make sense out of the cacophony we’re constantly bombarded with.

“Didn’t you hear what I meant?”

My native Arabic-speaking friends and I have a running joke in which we keep asking each other that question. Although we say it mostly in fun, it does have an element of truth. In much of our language, especially English with its quarter-million word vocabulary, meaning frequently gets lost in transmission.

Drawing from research into ancient texts and languages, we’ve gained insight into an elegant physiological design for expressing deep knowledge by assembling speech atoms (phonemes) using rules of syntactic and semantic aggregation, based on a small set of semantic primes. As we unveil those hidden rules, our capacity for machine understanding and knowledge representation dramatically improves.

The bottom line is we now have a far less complicated approach to semantic analysis that links narrative to the underlying knowledge, including the emotions behind it (for example, knowledge that is based solely on survival vs. the innate human desire for exploration and discovery). From counter-terrorism to good old-fashioned marketing, the payoff potential is huge.

Actions to take

Deep semantics is still in its early stages. Here are a few steps you can take to hop on board. Apply them to anything in need of greater clarity: business intelligence, intellectual property, marketing, public relations, strategy, workflow, policies and procedures, and, of course, knowledge sharing and transfer.

1. Expand your knowledge horizons and learn to see deep structure everywhere.

Come up with your own list of semantic primes. Which base emotions or values are at work in your enterprise? A sense of fulfillment? Destroying the competition? A sense of sharing or stewardship? How well are they reflected in your narrative?

Go beyond just looking at the words themselves. Pay attention to their basic “atomic” structure and valences. Do they attract or repel? Are they soft or percussive? Open or closed?

2. Spot trends sooner by engaging in the social discourse within and outside your organization.

Start building ontologies and putting them to use. Determine what rules of semantic aggregation are in play.

With practice, intent will become more transparent with less chance for deception. You’ll be able to better identify and correct false or negative memes that have been adversely impacting not only public perception about you and your enterprise, but also your own self-image.

3. Stop dumbing down and reverse the semantic loss that comes with it.

That doesn’t mean going back to the jargon-heavy styles that everyone rebelled against and that brought us to where we are today. But you should ignore the prevailing wisdom, which is to write at no higher than a seventh- or eighth-grade level. Rather, focus on keeping things simple and deep, as opposed to simple and shallow.

Identify the minimal set of semantic primes you need to deliver as clear and unambiguous a message as possible. Remember, the use of surface-level or shallow semantics quickly breaks down in complex, rapidly changing situations. Deep semantics allows you to navigate the complexity without being overwhelmed.

Here’s an added bonus step. What do most ancient scriptural texts have in common that makes them memorable? Stories, of course! So be sure to use them.

Be forewarned. This isn’t easy. It could even make your head hurt. But that might not be nearly as painful as continuing along a course marked by chaotic misunderstanding and not hearing what people actually meant.