The conventional linguistic theory only considers syntactic analyses. In this approach, the comprehension of syntactic elements obtained in the analysis process and their composed semantics as a whole for the sentence are overlooked. This may work to some extent for human comprehension of simple sentences. However, it is not only an incomplete process for human readers to deal with complex sentences, but also an inexplicit intermediate result for machines comprehension.
Therefore, a closed loop in sentence comprehension requires two additional steps beyond syntactic analysis, which are semantic analysis on the syntactic elements, and semantic synthesis with the terminal semantics of the syntactic elements [Wang and Berwick, 2012]. Modern technologies in computational linguistics and compiler theories have proven that these three steps are necessary for machine-based natural language processing.
Cognitive linguistics does not place a strong emphasis on syntax for determining semantic content. Consider the classic examples of:
1. The cat was on the table.
2. The cat was under the table.
These are syntactically identical but have significantly different meanings. The focus is on the different spatial relations signified in the prepositions.
The notion that we understand natural language primarily on the basis of syntax is flawed. When interacting with nonnative speakers of English, for example, a native speaker will to a significant degree ignore the syntax (for it is often incorrect) and focus on intention and meaning.
In cognitive semantics, the meaning of a sentence is assumed to be constructed on the basis of the linguistic material it contains. That is, words are understood as prompts for this process of meaning construction, and the latter is equated with conceptualization (inclusive of conceptual projections and the recruitment of background knowledge). Evans & Green, 2006, Cognitive Linguistics. An introduction, EUP, chapters 5 and 7 provide a useful synoptic overview of the nature of meaning from a CS perspective.