(1b) People who love apples. (2b) Fruit Mary loves.
(3b) Fruit John loves. (4b) Music lesson.
At this point it is intuitively clear that linvecel (on the other hand) in (1) is opposed to (1b), in (2) is opposed to (2b), and so on. This means that our encyclopedic competence has stored a semantic analysis of /invece/ where, after a general semantic marker of alternative, a contex-tual selection is recorded, such as «in the cases in which the textual topic is x, the expression under consideration marks an opposition to X» (Figure 2.10).
(contt…. /+x ) (preposition) ‘substitution with x’ /invece/ ‘alternative’
(conttopic X )(adverb) ‘opposition to X’
FIGURE2.10
2,3.2. Some attempts: registering frames and scripts
Nevertheless, there are cases in which such notions as contextual selec-tions are not enough to establish the possible textual insertion of given lexemes. Take an expression such as the following:
(5) John was sleeping when he was suddenly awakened. Somebody was tearing up the pillow.
I imagine that a computer fed with dictionary-like information would be able to understand what /to sleep/ and /pillow/ mean, but would be unable to establish what the relation is between John and the pillow (and which pillow?). Current research in Artificial Intelligence has elaborated the notion of frame (Minsky 1974; Winston 1977; Schank 1975, 1981;
van Dijk 1977): the addressee (be it a computer or a human being) is endowed with an enlarged encyclopedic competence which encompasses also a set of frames, or scripts, among which —for instance —are the frames ‘sleeping’ and ‘bedroom’. By resorting to this storage of compe-tence, the addressee knows that human beings usually sleep in bed-rooms and that bedrooms are furnished with beds, beds with pillows, and so on. By amalgamation of two or more frames, the addressee realizes that the pillow just mentioned can only be the one John was resting his head on.
To what extent can one assume that the frames, too, are elements of an encyclopedia-like componential analysis (an extremely rich one in-deed)? Do they belong to a sort of additional competence (that which is called overcoming in Eco 1976)? Probably the very notion of the encyclo-pedia has to be revised and reorganized according to different rates of social accessibility or of textual necessity. For instance, when reading John was sleeping and was dreaming of . . . , one does not need to have recourse to the frame ‘bedroom’ (even though one shares this piece of knowledge), and only the co-text leads one to blow up and to narcotize given sememes or frames (see Eco 1979).
The notion of frame (before being postulated by the empirical engineering of Artificial Intelligence scientists) was already indirectly advocated by Peirce. Take, for instance, his example of a possible defi-nition of lithium. According to a general dictionary (Webster’s New Col-legiate), lithium is «a soft silver-white element of the alkali metal group that is the lightest metal known and that is used esp. in nuclear reactions and metallurgy» (I have stressed the most evident elements of encyclopedic information).
According to a scientific handbook, lithium is definable as that ele-ment which has an atomic number 3, atomic weight 6.393, fusion point 108.5 degrees centigrade, boiling point 1330 degrees centigrade, density 0.53. This second definition, even though expressed in technical jargon, looks more similar to the following definition given by Peirce than it does to the one given by Webster:
If you look into a textbook of chemistry for a definition of lithium you may be told that it is that element whose atomic weight is 7 very nearly.
But if the author has a more logical mind he will tell you that if you search among minerals that are vitreous, translucent, gray or white, very hard, brittle, and insoluble for one which imparts a crimson tinge to an unluminous flame, this mineral being triturated with lime or witherite rats-bane, and then fused, can be partly dissolved in muriatic acid; and if this solution be evapo-rated, and the residue be extracted with sulphuric acid, and duly purified, it can be converted by ordinary methods into a chloride, which being obtained in the solid state, fused and electrolyzed with half a dozen powerful cells, will yield a globule of a pinkish silvery metal that will float on gasolene; and the material of that is a specimen of lithium.
The peculiarity of this definition — or rather this precept that is more serviceable than a definition — is that it tells you what the word lithium denotes by prescribing what you are to do in order to gain a perceptual acquaintance with the object of the world. (C. P. 2.330)
This ‘operational’ definition looks more like an informal frame than an encyclopedic description (for descriptions of this kind, see Charniak 1975> 1980). It satisfies the needs of an Instruktionssemantik. Gathering together the definition of Webster and that of Peirce, one is able to understand why, let us suppose, in a text a certain Professor Smith should need lithium for an atomic experiment and look for lime or witherite rats-bane. A textual assertion of this type undoubtedly elicits a lot of presuppositions, but these presuppositions are governed by preexisting frames.
Therefore, when a text theory aims to establish a «frame for frames» (Petofi 1976), it is attempting both to discover textual rules and to set up a more organized and comprehensive notion of code as encyclopedic knowledge.
Schank and Abelson (1977) attempt to represent not only the lexical meaning of a given expression but also all the connected forms of world knowledge that allow the interpreter to draw (from the utterance of a term or of a sentence made up with the analyzed terms) co-textual infer-ences. Thus (through the use of certain primitives representing funda-mental human operations, such as ATRANS, EJECT, INGEST, MOVE, and so on) Schank represents the verb to eat as an item susceptible of being inserted into contexts such as John ate a frog (Figure 2.11).
mouth1 John1
John1 INGEST frog1 mouth PART (JOHN1) MOVE
hand
у mouth
FIGURE2.11
An interesting semiotic problem (which has escaped the notice of Ar-tificial Intelligence theorists up to now) is the question of how to inter-pret in their turn not only the primitives expressed verbally but also the visual expressions of the diagram in Figure 2.10. In a general semiotic framework, a linguistic term can be interpreted by nonlinguistic inter-pretants, but even these interpretants are semiotic devices which must be in turn interpreted. In Eco (1976, 3.6.5) they have been studied as topo-sensitive vectors.
The project for an encyclopedic semantics is still in progress. Up to now we have been witnessing a series of alternative or complementary proposals. The structural semantics of Greimas (1966, 1979), with its notion of actant and of classemes or contextual semes, as well as with the idea of Narrative programs’, is encyclopedia-oriented; the early case grammar of Fillmore (1968), along with the more recent researches of Fillmore (I975, 1976a, 1976b, 1977, 1981), as well as the proposals by Bierwisch (1970, 1971), are encyclopedia-oriented. A new semantics in terms of encyclopedia and world knowledge (and phenomenological ex-perience) is evident in the last researches of Lakoff (1980) and Lakoff and Johnson (1980).
Since all of these representations concern coded sequences of actions, relationships between agents (actors or abstract actants), they can be re-traced together to the idea of frame.
2.3.3. Some attempts: stereotypes and commonsense knowledge
Undoubtedly, any encyclopedia-like semantics must blur the distinctions between analytic and synthetic properties, and in this sense nothing bet-ter has been said after and beyond Quine’s «Two Dogmas of Empiri-cism» (1951). Naturally, nothing prevents one from using analytical markers, provided one knows that they are shorthand devices used in order to include other analytic properties they entail and to summarize all the synthetic markers they name (in the same way as the names of genera, in a Porphyrian tree, name clusters of accidental properties or differentiae). In this sense, the proposal of Putnam (1975) is more than acceptable:
The normal form of description of the meaning of a word should be a finite sequence, or Vector’, whose components should certainly include the fol-lowing (it might be desirable to have other types of components as well): (1) the syntactic markers that apply to the word, e.g., ‘noun’; (2) the semantic markers that apply to the word, e.g., ‘animal’, ‘period of time’; (3) a de-scription of the additional feature of the stereotype, if any; (4) a description of the extension. The following convention is a part of this proposal: the components of the vector all represent a hypothesis about the individual speaker’s competence, except the extension. Thus the normal form description for ‘water’ might be, in part:
Syntactic markers mass noun, concrete
Semantic markers natural kind liquid
Stereotype colorless transparent tasteless thirst-quenching etc.
Extension H2O
(give or take impurities)
The idea of the stereotype represents probably the most fruitful sug-gestion coming from Putnam’s theory of language: according to our nor-mal competence it is doubtful whether tigers are ‘felidae’ or ‘felis tigris’ or ‘metazoa’, but they are undoubtedly ‘yellow’, ‘like a big cat’, and ‘black striped’.
Semantic markers are pseudoanalytic shorthand labels, since there should be a stereotype also for ‘liquid’ and perhaps for ‘natural’.
As for the extension, Putnam’s proposal is dependent on his sharing of Kripke’s theory of rigid designation (Kripke 1972). It is hard to say whether Putnam’s idea of rigid designation really corresponds to that of Kripke; Putnam thinks of the something rigidly designated by a linguis-tic term as an essence that can be defined in scientific terms. However,
such an expression as Ή2 Ο’ is composed by chemical symbols, which, in turn correspond
