Introduction The term frame semantics refers to a wide variety of approaches to the systematic description of natural language meanings. The one common feature of all these approaches – which, however, does not sufficiently distinguish frame semantics from other frameworks of semantic description – is the following slogan due to Charles Fillmore (AAA): Meanings are relativities to scenes. According to this slogan meanings have internal structure which is determined relative to a background frame or a scene. The easiest way to understand this thesis is by way of example.
The following one is from Fillmore (ICC): Suppose that two identical twins Mark and Mike are both in a hospital sitting on the edge of their beds in exactly the same position. When a nurse walks by Marks room, she says: I see that Mark is able to sit up now, and when she walks by Mike’s room she remarks: I see that Mike is able to sit down now. Drawing on what we know about hospitals – our hospital background scenes or frames – we will interpret the two remarks of the nurse rather differently, thereby relativities the meanings of her remarks to the relevant scenes.
Another often cited example of Fillmore (ICC) clearly mistreating the above thesis is the difference in meaning between the following two sentences: (1) I spent three hours on land this afternoon. (2) I spent three hours on the ground this afternoon. The background scene for the first sentence is a sea voyage while the second sentence refers to an interruption of an air travel. This illustrates Fillmore use of the term frame as an idealization of a coherent individuate perception, memory, experience, action, or object Fillmore (ICC).
In order to understand frame semantics, it is helpful to begin with a brief history. From here we will turn to an overview of the most important theoretical concepts. After this the relationship of frame semantics to one specific version of Construction Grammar will be introduced and some examples will be analyses. The paper will end with a short summary of applications of frame semantics and a note on formalization. Usually frame semantics is taken to be a very informal approach to meaning, but nevertheless some approaches relating frame semantics to Formal Semantics exist.
History There are at least two historical roots of frame semantics; the first is linguistic Syntax ND Semantics, especially Fillmore case grammar, the second is Artificial Intelligence (AY) and the notion of frame introduced by M. Minsk (1975) in this field of study. Which identifies (at least) the participants of the scene and thus the arguments of predicates and sentences describing the scene. In order to understand a sentence the language user is supposed to have mental access to such schematic scenes. The other historical root of frame semantics is more difficult to describe.
It relates to the notion of frame-based systems of knowledge representations in AY. This is a giggly structured approach to knowledge representation which collects together information about particular objects and events and arranges them into a taxonomic hierarchy familiar from biological taxonomies. However, the specific formalism suggested in the above mentioned paper by Minsk was not considered successful in Some Basic Theoretical Principles The central theoretical concepts characterizing frame semantics are due to C. Fillmore and did not change much since his first writings on this approach.
In order to explain the most important notions of frame semantics let us briefly consider a happily example of a frame, the commercial transaction frame which demonstrates the origin of frame semantics from Fillmore case frames as well. In this case the concept frame is applied to verbs like buy with the intention to represent the relationships between syntax and semantics. BUYER buy GOODS (PRICE) object subject (SELLER) from from Pete for $ 10 Angela bought Eddy them Penny a bicycle for $ 1 from Stephen The verb buy according to the above table requires obligatorily a buyer, goods and optionally a seller and a price.
Verbs with related meanings such as sell are expected o have the same meaning slots but in a syntactically different order. This clearly shows the relation to Fillmore case frames. Combining these frames results in the commercial transaction frame about which the following table provides partial information: VERB SELLER MONEY PLACE sell to cost indirect at spend 2 Of course the PLACE-feature Just marks the beginning of an open ended list, since every event in the above table can be further specified – for instance with respect to time. Moreover the collection of frames forms an ordered structure.
For instance the immemorial transaction frame itself is part of the more general transaction frame prototypically expressed by the transitive verb give. This indicates that the system of dependencies between frames forms an intricate hierarchical structure. The concept Prototype is one of the most important concepts of frame semantics. Frames should be understood as prototypical descriptions of scenes. A prototype has the advantage that it does not have to cover all possible aspects of the meaning of a phrase; in other words a prototype does not have to provide necessary and sufficient conditions for the correct use of a phrase.
Fillmore (Bibb) illustrates the use of prototypes within frame semantics by an analysis of the concept widow. The word widow is specified with respect to a background scene in which people marry as adults, they marry one person, their lives are affected by their partner’s death and perhaps other properties. The advantage of a theory of meaning based on the prototype concept compared to a theory which insists on stating necessary and about certain boundary conditions; that is it does not have to provide answers for questions like WSDL you call a woman a widow who has lost two of her three cabanas but who had one living one left?
Fillmore (Bibb). In a case like this whether the noun widow applies or not is unclear since certain properties of the background frame for this concept are missing. However, the concept prototype is not unproblematic either. Note that Fillmore does not use this concept with respect to words but with respect to frames or scenes. Some words like bird certainly have prototypes but others may not have a corresponding prototype. What is a prototypical vegetable for instance or a prototype corresponding to the adjective small? Moreover, applications of prototype theory often involve two different measures for category membership.
A penguin, for example, is certainly not a prototypical bird, but nobody hesitates to Judge it as a bird. The other measure of category membership is typically used in the analysis of vague predicates, for instance color adjectives. It may sometimes be hard or even impossible to assign a given object to the category of pink or red entities . Another central notion within frame semantics is the concept profiling. Linebacker (1987) uses the example of hypotenuse for explaining this concept. One can easily draw a mental stricture of the concept hypotenuse.
The interesting question concerning this mental picture is this: Can you imagine what a hypotenuse is without imagining the whole right triangle? The answer is clearly: no. The triangle and the plane it is included in is a frame, and the term hypotenuse and right triangle are interpreted with respect to this frame but they profile different parts of the frame. The following example taken from Goldberg (1995) illustrates lexical profiling of participants. Consider the following differences between the closely related verbs rob and steal. A. Jesse robbed the rich (of all their money). B. Jesse robbed a million dollars (from the rich). A. Jesse stole money (from the rich). 3 b. *Jesse stole the rich (of money). These distributional facts can be explained by a semantic difference in profiling. In the case of rob the victim and the agent (the thief) are profiled, in the case of steal the agent and the valuables are profiled. Representing profiled participants in boldface, Goldberg proposes the following argument structure for rob versus steal: rob steal However, Goldberg main concern is with constructions for which she uses frame meantime in order to provide highly structured rich meanings for them.
What are constructions ? Here is A. Goldberg definition. A construction is defined to be a pairing of form with meaning/use such that some aspects of the form or some aspect of the meaning/use is not strictly predictable from the component parts or from other constructions already established to exist in the language Goldberg (1995). There is no doubt that constructions exist. Morphemes for instance satisfy Goldberg definition. But do constructions different from morphemes exist? This is of course hat defendants of construction grammar try to show.
Here we will take the existence of constructions other than morphemes simply for granted. Consider the following examples: Margaret baked Peter some cookies. Martin sneezed the napkin off the nightstand. The peculiarity of example (5) is due to the fact that the verb bake, which normally has two arguments, is used with three arguments here. Particular as this sentence is we nevertheless can make sense of it. Margaret baked some cookies with the intention to give them to Peter. Note that this interpretation helps us to make sense f the recipient role which is not provided by the verb bake; I. E. E think of this sentence as an instance of the transitive construction of which a more standard example is: (7) John gave Mary a present. The crucial claim of construction grammar is that this is not due to different basic meanings of the verb bake but due to the integration of this verb plus its meaning into the transitive construction which has a meaning of its own. Therefore construction grammar distinguishes the semantics of argument structure constructions from the semantics of the verbs which instantiate them. An advantage f this approach is that it accounts for novel uses of verbs in specific constructions.
In (6) the intransitive verbs sneeze has to be integrated into the caused motion construction and therefore is forced to be interpreted as some kind of action. Both verbs and constructions are associated with frame semantic meanings. However, in contrast to the rich frame semantic representations of verbs the basic construction are associated with a more abstract semantics. These basic constructions 4 and their frames are supposed to be independent of a particular language. They are the content of Goldberg Scene Encoding Hypothesis.
Scene Encoding Hypothesis: Constructions which correspond to basic sentence types encode as their central senses event types that are basic to human experience Goldberg (1995). Applications Frame semantics has a wide range of applications reaching reaching from subfields of linguistic theorizing such as Morphology to Typology, Discourse Analysis, and Language Acquisition. However, the central and most successful application seems to be (computational) lexicography. In a frame based lexicon the frame accounts for related senses of a single word and its semantic relations to other words.
A frame eased lexicon therefore offers more comprehensive information than the traditional lexicon. An example is Petcock (1986) which studies the vocabulary of the body frame in Modern Hebrew. An example of computational lexicography is the Frameset-System (see Boas (2002)). Formalization Although frame semantics does not lend itself easily to formalization there is an early approach by Grown (1983) in which basic insights of frame semantics were formalized by LISP-Ice notations in combination with situation semantics.