Natural Language Processing and Computational Linguistics 2

2The Sphere of Semantics

2.1. Combinatorial semantics

Far from the theoretical or psycholinguistic controversies about the role and manner in which humans combine different resources of linguistic information to construct meaning in a sentence, it is generally accepted that the lexicon and syntax play a non-negligible role in this process¹. Because these two sources of information are not always sufficient, high-level semantic constraints are indispensable for distinguishing different possible interpretations of the same syntactic structure. For example, the subject and the object of a sentence can play different semantic roles. Sometimes, the subject is the direct cause of an event that is voluntary or not (e.g. John writes), or the indirect or unconscious cause (e.g. it is raining). Sometimes, it is the subject that is affected by the action expressed by the verb (as in, John dies).

2.1.1. Interpretive semantics

At the start of the 1960s, Fodor and Katz [FOD 64] proposed enriching syntactic structures constructed in the form of trees with a semantic analysis.

Interpretive semantics establishes a clear distinction between aspects of interpretation that are founded on linguistic knowledge and aspects of interpretation that are derived from knowledge about the world. According to this theory, it is this distinction that makes it possible to draw the boundary between semantics and pragmatics. It assumes that semantics must account for all of the possible interpretations of ...