The Formalization Of Pragmatic Inference

Semantics traditionally can be divided into Lexical Semantics and Compositional Semantics. Lexical Semantics means the contributions of individual words to discourse contents, and Compositional Semantics holds that the way meanings of individual words compose together to yield sentence contents. But people often mean more than what is said. The interpretation of utterance only depending on grammar and sentence structure is insufficient. Pragmatic inference plays an important role in the process of determining meaning, without it, some utterances are incoherent or"make no sense". The goal of pragmatics is to explain how the gap between sentence meaning and speaker's meaning is bridged.The formalization of pragmatic inference is a formal approach on pragmatic inference. The formal approach based on logic and philosophy features in strictness, precise, the formal approach present a strict theory, reveal the nature of language, especially in nowadays, the approach is very popular with the development of the computer. The formal research of language is getting more and more important.The formal approach on natural language can be traced back to the true semantic theory of Tarski in which basic thought and method about formal semantics were formed, but his theory only apply to formal language and artificial language in logical research, it can not be applied to natural language directly. Montague made a development of Tarski'truth semantic theory and applied it to natural language, formed a formal semantic theory of natural language: Montague Grammar. Montague pragmatics is actually an extension from formal semantics to formal pragmatics. It can be said that extension towards pragmatic issues based on model theory has been a basic development idea of formal pragmatics. But Montague Grammar also exposes its shortages soon, many natural language phenomena can not be explained reasonably, such as in the aspects of personal pronoun anaphora and quantifier domain, etc. Though it is a particle problem compared with Montague Grammar's unified program and theory on the various language expressions, it reveals that there are some fundamental characteristics behind the natural language and the past theories have deficiencies on awareness of these characteristics. Both the Tarski's Theory and Montague Grammar are static approach.DRT (Discourse Representation Theory) and SDRT (Segmented Discourse Representation Theory) are two versions of dynamic approaches. DRT arose mainly aiming at resolving"indefinite descriptions"and"donkey anaphora", DRT regards meaning as a psychological phenomenon, an externalization of human thought. DRT are interested not only in the truth conditions of sentences, but also in how sentences are interpreted by hearers. DRT theorists assume the language user interpret sentences by constructing abstract semantic structures, called DRS, the representation of discourse are mental representations. They rely on model-theoretic semantics but also on the meaning grasped by the hearer on hearing an utterance. These structures are results of the application of rules which are called DRS Construction Rules. The novelty is that these rules also take into consideration the information from the previously constructed DRSs, reflecting the fact that interpretation of discourse is incremental and that discourse is cohesive. This ability to account for changing context is the essential property of dynamic semantics. An interpretation is obtained for the first sentence, and then this interpretation serves as context for the interpretation of the next sentence.DRT is successful to explain"indefinite descriptions"and"donkey anaphora", but there are many natural language phenomenon can not be formalized reasonably by DRT, such as bridging anaphora and conversational implicatures.SDRT is based on DRT and more complicated than DRT. SDRT absorbed DRT's expressive pattern and introduced rhetorical relations into discourse interpretation. Compared with DRT, SDRT have made great breakthrough in thoughts, theories and techniques. In this sense, SDRT is not a branch of DRT, but a new semantic theory over DRT.SDRT holds that sentences in discourse have different relations though they look like strings, the relationship between sentences are not equal, some sentences connected closer than others, so discourse segment comes into being. SDRT argues that the meaning of discourse depends upon and interacts with its rhetorical structure, which links the meaning of the sentences together in a coherent whole. In other words, interpreting discourse exploits domain knowledge, cognitive state, lexical meaning, compositional meaning and rhetorical relations, and SDRT is a theory of discourse interpretation account for how these knowledge sources interact. SDRT not only integrates an account of rhetorical structure with dynamic semantics, it also gives a formal account of pragmatics and semantics that goes beyond other approaches.SDRT accounts for several interrelated discourse phenomena: the resolution of anaphora of various kinds including bridging anaphora; word sense disambiguation and conversational implicatures, etc. For the limitation of my logical knowledge, this thesis only accounts for some anaphora phenomena.As any other theories, SDRT has its own limitations, Such as SDRT is still not give detailed analysis on most of the discourse relations that are used in daily life. There is no investigating to get wider empirical coverage, etc.This thesis firstly gives an simple overview on traditional static formal approach, and introduces the basic logical knowledge, mainly focuses on propositional logic and the first order predicate logic. Based on all of these preliminaries, this thesis presents two versions of dynamic semantics, they are DRT and SDRT. This thesis focuses on the basic thought and theoretical frame of SDRT, and applies the method of SDRT to formalize bridging inference, and this point is the novelty of this thesis as well.