Article Preview
TopIntroduction And Outline
Artificial intelligence (AI) systems seek to simulate some aspects of human cognition (Sowa, 2011). Since cognition appears to rely heavily on the use of words, researchers are trying to understand the relationship between word meanings and cognitive processes (Sowa, 2011). There is evidence that the meanings of words and expressions are connected with cognitive mental structures that can be represented as graphs (Sowa, 2011). These cognitive structures are commonly known as conceptual structures (Sowa, 2011). Some refer to them as frames (Petruck, 1996; Fillmore, 1982).
Unfortunately, in their search for cognitive structures, researchers have only been relying on the study of the meanings of whole morphemes (the smallest grammatical units: word roots, pronouns, articles, prepositions, etc.), without examining the semantic contribution of the individual sounds that make up those morphemes to the meanings of the morphemes. This is obvious, for example, from Sowa’s discussion of the expression “a cat sitting on a mat” (around Figure 1 and Figure 2 in: Sowa, 2011), and Petruck’s and Fillmore’s discussions of frames (Petruck, 1996; Fillmore, 1982).
Figure 1. How the abstract symbol table was found
Figure 2. Readware’s best performance at TREC 8
This approach to semantics is based on the doctrine that individual sounds don’t contribute to the meanings of morphemes at all, a doctrine that was established by the linguist Saussure after his extended search for sound meanings ended in failure (Saussure, 1916).
We call the type of semantics that studies morpheme meanings but ignores the contribution of individual sounds to morpheme meanings shallow semantics. Deep semantics, on the other hand, examines the contribution of individual sounds to morpheme meanings. We believe that deep semantics research is more likely than shallow semantics research to uncover the real cognitive structures that people use.
The author has studied deep semantics for three decades. We will introduce some results of this research. We will first explain why the author succeeded in finding some consistent rules for the contributions of individual sounds to morpheme meanings after Saussure and others despaired of ever finding such rules. Then, we will introduce the basic findings of deep semantics and the nature of the cognitive structures that it offers. Next, we will document the implementation and practical validation of the findings of deep semantics.
As it turns out, deep semantics also offers an ontology maker: an algorithm for creating ontologies in the original philosophical sense, i.e. integrated theoretical frameworks that explain the functions and elements of the domains of the real world (Smith, 2003 & 2004). A theory of emotions and a consumer choice theory were created using the ontology maker (Adi, 2009; Adi & Nevers, 2014). We will demonstrate the ontology maker by using it in an attempt to start building an ontology of the domain of cognition itself.
Finally, we will perform a brief comparison between some cognitive structures based on shallow semantics and some cognitive structures based on deep semantics, and then end the paper with a conclusion.
TopA Brief History Of Deep Semantics Research
Morphemes consist of individual sounds, and it is logical to assume that individual sounds play a role in determining the meanings of morphemes.