| << Prev | - Up - | Next >> |
An example for using a theorem prover in NLP.
Conversational Maxims
Now that we have a computational method for solving inference problems, let's look at a case where we can apply it in semantic interpretation. We shall use inference to check whether a speaker obeys the conversational maxim s in his utterance. The notion of conversational maxims was introduced by H.P. Grice in 1975. He postulates a set of constraints on discourses, which he formulates as maxims for the speaker. These maxims characterize discourse as rational cooperative activity. The hearer can assume that the speaker follows these maxims, and on this assumption can draw inferences to the intended interpretation of the discourse: Often if one of multiple reading violates a maxim, then it simply cannot be the the intended one.
Conversational Implicatures
In other cases a violation allows to infer ``backwards'' to an intention or assumption on the side of the speaker. Intentions and assumptions that can be inferred from violations of the conversational maxims are called conversational implicature s.
Be cooperative!
Grice assumes that participants in a discourse follow a general cooperative principle . This principle leads to more specific submaxims, falling into one of four categories:
Try to make your contribution one that is true.
Make your contribution as informative as is required.
Be relevant.
Be perspicuous
Generally, Grice's maxims are viewed as pragmatic in nature. As regards the maxims of manner and relation, it may indeed not be easy to see how being relevant or being perspicuous could be defined solely in semantic terms, without reference to more general factors such as e.g. the intentions, mutual knowledge or the sociolect of speakers/hearers. In contrast, we can get a grip on the first two maxims without having to tackle all of the complexities of pragmatics, if we use inference techniques on our semantic representations.
| << Prev | - Up - | Next >> |