For the slot filling task, the input is the sentence consisting of a sequence of words, L, and the output is a sequence of slot/concept IDs, S, one for each word. In the statistical SLU systems, the task is often formalized as a pattern recognition problem: Given the word sequence L, the goal of SLU is to find the semantic representation of the slot sequence 푆that has the maximum a posterioriprobability 푃(푆|퐿).

WALTER 1. ONG, SJ

it occurred to me that cheer andoptimism aren’t really my thing

or instance, young animals are endearing, even when the adults of the species are not; so the child may feel encouraged to derive not only the obvious contextual implication that he is dirty, but also the further contextual implication that he is, nevertheless, endearing.

The greater effort imposed indicates that greater effect is intended. By uttering (13), the speaker thus encourages the hearer to look for a range of further contextual implications not shared by (14), and to assume that within this range there are some that she is prepared to endors

f we are right, loose uses are non-literal uses in the sense described above: they are based on resemblance relations among representations, and involve interpretive rather than descriptive dimensions of language use

The qualification 'near Paris' demands some processing effort, which, given the presumption of relevance, should be offset by some cognitive effect

Hence, Peter is entitled to assume that Marie intended him to interpret her utterance in this way, which is consistent with the principle of relevance

n a nutshell, for an utterance to be understood, it must have one and only one interpretation consistent with the fact that the speaker intended it to seem relevant to the hearer-adequately relevant on the effect side and maximally relevant on the effort side

conveyed, but not asserted

Speakers may try hard or not at all to be relevant to their audience

In such a process, the hearer is taking a large share of the responsibility for the conclusions he arrives at. As a result, different hearers with different background knowledge and different imaginations will follow somewhat different routes.

We want to argue that they differ not in kind but only in degree of looseness

We claim that interests are simply by-products of the general search for relevance: as a result of our cognitive history, some topics in our memory are richer in information and, either temporarily or permanently, more accessible than others, so that information relating to them is likely to produce greater effect for less effort, i.e. be more relevant as defined.

On this approach, the claim that a man with one hair is bald is just as false as the claim that a man with a full head of hair is bald. What distinguishes them is not the fact that one is true and the other false, but the fact that one is an acceptable loose use because many of its logical and contextual implications are true, whereas the other is unacceptable since a hearer would be able to derive from it virtually no true descriptive information about the state of affairs it purports to represent

nd implication (9), which is contextually implied by both

Marie's first answer, 'I live in Paris', is effective enough to convey just what she wants; it may be more effective than the literally true second answer, 'I live near Paris'.

ny object in the world can, in principle, be used to represent any other object that it resembles. For instance, a piece of rope can be used to represent a snake which it resembles in shape

but also calculable

: it is an exceptionless generalisation about human communicative behaviour.

Any utterance addressed to someone automatically conveys a presumption of its own relevance. This fact, we call the principle of relevanc

We claim that humans automatically aim at maximal relevance, i.e. maximal cognitive effect for minimal processing effort.

For instance, when I draw you a diagram of how to get to my house, you do not infer that I intend you to travel across white paper, in two dimensions, past landmarks clearly labelled CHURCH and NEWSPAPER SHOP, a distance of 8 inches from door to door. You have to make some assumption about which properties of the representation carry over to the original.

Every utterance used in verbal communication interpretively represents a thought entertained by the speaker-the very thought that the speaker wants to communicate. That much the hearer is entitled to expect; that much is necessary for verbal communication to be possible at all. However, the hearer is not invariably entitled to expect a literal interpretation of the speaker's thought, nor is such an interpretation always necessary for successful communi- cation to take place. A less-than-literal interpretation of the speaker's thought may be good enough: may indeed be better on some occasions than a strictly literal one.

Let us say that when one representation is interpretively used to represent another, all of whose implications it shares, it is a literal interpretation of that other representation. On this account, literalness is just a limiting case of interpretive resemblance.

By our definition, propositions (4) and (5) resemble one another more in context (6a-b) than in context (7a-b): in (6a-b) they share implication (8), which is contexually implied by (4) and analytically implied by (5), and implication (9), which is contextually implied by both; whereas in (7a-b), (4) and (5) share no implications at all. This seems to match our intuitions, insofar as intuitions are possible given the artificiality of the example.

We are thus defining interpretive resemblance as a context- dependent notion: two propositions P and Qmay resemble one another closely in one context and less closely or not at all in another context.

interpretati

r descripti

Any object in the world can, in principle, be used to represent any other object that it resembles. For instance, a piece of rope can be used to represent a snake which it resembles in shape. An utterance can be used to represent another utterance which it resembles in meaning-either closely, as in the case of a paraphrase or translation, or more distantly, as in the case of a summary. Generally speaking, an utterance can be used to represent any representation which it resembles in content, whether a public representation such as another utterance, or a mental representation such as a thought.

Rather, she intends him to entertain the proposition(s) most salient in her mind and to construct around it (or them) a complex thought which merely bears some similarity to her own. For instance, the mother wants the child to realise quite clearly that she thinks he is dirty, and to get at least an inkling of her accompanying thoughts

In her modest way, the mother who calls her child a piglet achieves some unparaphrasable effects: for instance, she seems more indulgent than if she had called him a dirty child.

Such views are no longer considered adequate to account for other cognitive abilities, but are still called upon, for want of any alternative, when it comes to explaining what is evoked by a metonymy, a synecdoche, a metaphor, or an irony. No other explanation is given of how figurative interpretations are recovered. Grice's account merely adds an inferential step of confirmation to these mysteriously retrieved figurative interpre- tations.

The initial implausibility of any hypothetical rule of literal truthfulness might be overlooked if the appeal to such a rule had useful theoretical consequences; if it helped to explain how not only literal talk, but also loose talk and metaphor are understood. But in this respect, modern accounts are neither essentially different from, nor superior to, classical rhetorical accounts

One generally accepted answer is that there is a rule (or norm, or principle, or maxim, or convention, or presumption) of literal truthfulness whereby the utterer of a declarative sentence, in expressing a certain proposition, automatically vouches for its truth (similar rules of literal commitment can be formulated for non-declarative utterances)

Literal talk, loose talk and metaphorical talk are often seen as different in kind. We want to argue that they differ not in kind but only in degree of looseness, and that they are understood in essentially the same way.

technology

or the task was to give the meaning of all expressions in a certain infinite set on the basis of the meaning of the parts ;

rege's theor

ost philosophers of language, and recently even by some linguists,

. There is only one world, the “real”world:

Logic, I should maintain, must no moreadmit a unicorn than zoology can; for logic is concerned with thereal world just as truly as zoology, though with its more abstract andgeneral features.

Who did you meet?” “I met a man.” “That is a very indefinitedescription.” We are therefore not departing from usage in our termi-nology. Our question is: What do I really assert when I assert “I meta man”? Let us assume, for the moment, that my assertion is true,and that in fact I met Jones. It is clear that what I assert isnot“I metJones.” I may say “I met a man, but it was not Jones”; in that case,though I lie, I do not contradict myself, as I should do if when I sayI met aj168man I really mean that I met Jones. It is clear also that theperson to whom I am speaking can understand what I say, even if heis a foreigner and has never heard of Jones