Conversational Natural Language English Language Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.


A basic underlying assumption we make when we speak to one another is that we are trying to cooperate with one another to construct meaningful conversations. This assumption is known as the Cooperative Principle. As stated in H. P. Grice's "Logic and Conversation" (1975):

"Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged"

In other words, we as speakers try to contribute meaningful, productive utterances to further the conversation. It then follows that, as listeners, we assume that our conversational partners are doing the same.

Conversational Natural Language

Frederking, (1996) argued in his paper which titled as Grice's Maxims: "Do the Right Thing" that Grice's maxims are hopelessly vague, and that while his cooperative principle may be useful at a high level of theoritical analysis, it should not be directly implemented in computational natural language systems.

Grice's maxims are hopelessly vague, ambigiuos , and harmful, Frederking, (1996) explained that, because they structure and shape a misleading taxonomy. While his cooperative principle may be practical at a high stage of theoretical analysis, it too is vague, and could not be straightforwardly implemented in computational natural language systems. Answers are suggested to a number of this symposiums topics based on this situation. Examples are existed to present that the maxims are too vague and too general. In fact, they are not really used by computational systems that claim to be based on them. The historical origin of the maxims in Kant's philosophy are exposed. A contrast is made with Relevance Theory, which provides a better approach to the similar phenomena. I agree with Frederking who suggested that it might be too early to expect to find such broad principles in the history of computational linguistics.

From this point of view, there are clear answers to a number of questions that Frederking, (1996) addressed them as follow:

Is the notion of conversational implicature still useful?

What role if any do grice's maxims and Cooperative Principle still play in computational and formal approaches?

What distinguishes conversational implicatures from other defeasible inferences in discourse (e.g. default inferences in text understanding)?

Is relevance a well-defined notion?

The notion of conversational implicature, and Cooperative Principle as Frederking, (1996) explained have been useful and important to some researches in thinking about how language works in real use. But however useful they are for giuding a researcher thinking, they are not useful as an actual part of an implemintation. The maxims, on the other hand, do not play useful role at all in any computational or formal approaches, even at a theoritical level. In fact they are harmful because they shape a misleading taxonomy. Frederking, (1996) claimed in aswering question three that, the only thing that distinguishes conversational implicatures as a class is the fact that they can be seen as examples of Grice's principles. In other word, the categories he uses have no prodictive or discriptive power. This is not to declare that specific subclasses (such as scalar implicatures) do not have useful distinguishing aspect; only that the Gricean level of discription is misleading. Finally, Frederking, (1996) proposed that Relevance is not a well-defined notion, he explained, like the other maxims, Grice's Relevance is a broad, general declaration that is obviously true at some levels, but it is still ambiguous to be used directly in computational systems. He believed that it is currently an open question whether some other approach to relevance (such as Relevance Theory) could be amenable to specific or exact definition.

Frederking, (1996) added that most models have been focused on one classes of conversational implicature, and asked the question, what problems would arise in integrating them? It is meaningless to judge integrating several classes of conversational implicature, based on Grice's taxonomy. His taxonomy is not of a sufficiently real nature to be successfully in order to apply to real implementaions. There are major issues in intergrating different kinds of inference mechanisms in conversation, but Grice's categories are unrelated to these issues.

In short Frederking, (1996) claimed that Grice's Maxims are similar to the maxim "Do the Right Thinh" which any correcltly working natural language system can be said to implement.

Grice's Maxims Considered Harmful

In explaining "Grice's maxims considered harmful" as the the auther proposed, he explained that several researchers have tried to apply Grice's maxims in one way or another. For example, Gazdar, (1979) and Hirschberg, (1985). Because of the vagueness of the maxims, This seems to be impposible to apply, therefore they normally have implemented something more logically and practically, and then claimed it was "Gricean". Furthermore, researchers have tried to use Grice's maxims to explain specific phenomena or system. For example, Dale and Reiter, (1995), Josh et al, (1984), and Passonneau, (1995). Because the maxims have the form or the shape of a taxonomy, they guide other researchers to think that the maxims taxonomize the space of conversational implicatures in some useful fashion. But using the maxims even in this way is counter-productive, because they are too vague, and usually extend beyond while applied to real examples of conversational implicature. From my point of view, i agree with the auther in this because using the maxims in that way could likely to lead to confusion more than enlightenment.

Frederking added that it is useless to discuss whether a particular phenomenon like irony is based on the flouting of Relevance or Quality as Levinson, (1983) does. Irony is a phenomenon that fits quite comfortably into both notions. It flouts both at once, and perhaps Manner, too. The desire to decide which maxim irony flouts is based on the false impression that there is some kind of significant difference between implicatures that fits into one category and those that fit in the other.

The maxims not only divide discourse phenomena up badly; they also group them together badly. Scalar and clausal quantity implicature are a good examples (Gazdar, 1979 & Hirschberg, 1985). These have been described as subtypes of Quantity implicatures. But there does not seem to Frederking, (1996) to be any good reason to believe that these are two subclasses of the same phenomenon. Clasual imlicature typically occurs when an embedded proposition is neither affirmed nor denied by the full utterance. For example, the utterence of "If Sara sees me then she will tell Suzan" Implicates that the speaker does not know whether Sara will see her implicates that the speaker does not know whether Sara will see her. The standard explanation of this is that, base on the Cooperative Principle, if the speaker knew whether the first clause was true or false, she should have said so. On the other hand, Scalar implicature is based on the existence of sets of terms that have some salient partial ordering in degree of informativeness. So, the utterance of "Dan ate some of the eggs" implicates that the speaker does not know that Dan ate all of the eggs. Again, the standared explanation of this relies on Quantity, that the speaker should have said so if he knew.

Obviously, the pevious explanations are similar in character, and unfortunately, speakers often provide less or more information than is necessary and because of that, the generalization made by the maxims is not vaild.

Trying to Make Use of a Misleading Taxonomy

Frederking, (1996) proposed in trying to make use of a misleading taxonomy that the maxims are clearly true in some sense and this coupled with thrie vagueness has allowd large number of researchers to read into them all sorts of specific true interpretations rather than treating them as maxims, as Grice's name for them suggests (although Grice apparently did intend for them to be applied rigorously). When we examine what acually exists in specific systems that are claimed to fulfill one or more of the maxims, we find much more specific mechanisms that apply to much more phenomena and only bear a very tenuous connection to the maxims. The clarity of these individual phenomena and rules despite any remaning controversies is in sharp contrast tothe haze of confusion surrounding the maxims. The fact that the researchers often point out major problems with the maxims, it is difficult to understand the widespread, seemingly willful refusal to realize that the maxims simply are not correct. Hirschberg, (1985) rededined Quality very narrowlly and indicated that Quantity, Relevance, and Manner could not be defined precisely. She them wrote logical formula containing the maxims as if they were rigorously definable. The basic problem is vagueness; Grice's maxims are loaded with terms that are ill-defined, such as "as informative as required". Another example is Levinson, 1983) of assymetric "and". The fact that "and" could be used to mean "and then", and this is not a lexical ambiguity seems to Frederking to have been clearly established at this point. But this fact hardly makes the "Be orderly" submaxim of Manner a generally useful computational rule. There are contexts in which it is quite acceptable to describe events out of order. For example if one is telling a story and says "George says he likes Sara, and Hari walks out the door" there is a clear implication of sequentiality. However, if someone says instead "George says he likes Sara, and Hari says he likes Sara" there is no implication of sequentiality.

Frederking, (1996) depeneded on a better example of the tendecy to apologize for Grice by the discussion of the generating of referring expression that given by Dale and Reiter, (1995). They gave a typical example of obeying Quantity: A speaker who says "look at the pit bull" rather than "look at the dog" implicates that the type of dog is important, perhaps because it is more dangerous. Another example at the level of the maxims that does not obey Quantity: imagine there is a room containing just alligators. Native English speakers would usually refers to the largest one as "the largest alligator" rather than "the largest animal" or "the largest thing" in a simple way because "alligators" is the unmarked level of description which English often uses. So, why does this not generate a conversational implicature? A simple answer, according to the maxim of Quantity, a speaker generating clearly superflous information should cause the hearer to produce implicature like in the previous example.