Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In Proceedings of 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....(3) a. The rabbit in the hat b. Remove the rabbit from the hat. Thus an important shortcoming of this rst type of approach is that it fails to address the question of when and how some of the identifying properties may be left implicit when realising the de nite NP. To remedy this shortcoming, Stone and Webber (1998) argue that a generation architecture is needed which interleaves rather than pipeline contextual reasoning (e.g. the computation of a uniquely identifying set of properties) and sentence realisation (e.g. the surface realisation of a de nite NP) Speci cally, they show that the SPUD ....

....set of properties) and sentence realisation (e.g. the surface realisation of a de nite NP) Speci cally, they show that the SPUD architecture presented in Stone and Doran s (1997) appropriately captures the type of textual economy illustrated by the above examples. In this paper, we follow Stone and Webber (1998) and adopt a SPUDlike architecture to generate indirect anaphors. This architecture consists of three main components: A Lexicalised Tree Adjoining Grammar (LTAG, Joshi and Schabes 1997) A chart based generation algorithm (Kay 1996; Carroll et al. 1999) which uses the LTAG to build a ....

[Article contains additional citation context not shown here]

Stone, M. and B. Webber (1998). Textual Economy Through Close Coupling of Syntax and Semantics. In Proceedings of INLG 1998, pp. 178{ 187.

....re evoke the appropriate discourse entity. However, a goal directed view of sentence generation suggests that speakers can attempt to satisfy multiple goals with each utterance [Appelt, 1985] and that a single linguistic form can opportunistically contribute to the satisfaction of multiple goals [Stone and Webber, 1998]. The possibility that goals besides identification could influence the content of a nominal expression 1 have not been fully addressed in computational work on generating nominal expressions. The many one mapping of goals to linguistic forms is more generally referred to as overloading ....

....not been fully addressed in computational work on generating nominal expressions. The many one mapping of goals to linguistic forms is more generally referred to as overloading intentions [Pollack, 1991] Overloading can involve tradeoff across linguistic levels [Di Eugenio and Webber, 1996, Stone and Webber, 1998] For example, an intention which is achieved by complicating a form at the semantic level may allow the speaker to simplify at the syntactic level by omitting important information [Stone and Webber, 1998] Although we have learned that overloading is natural and perhaps even necessary, we have ....

[Article contains additional citation context not shown here]

Stone, M. and Webber, B. (1998). Textual economy through close coupling of syntax and semantics. In Proceedings of 1998 International Workshop on Natural Language Generation, Niagra-on-the-Lake, Canada.

....in that it would not be useful only for Natural Language purposes. It would represent all aspects of actions needed for performance and therefore it would support the generation of effective instructions. The Natural Language generator SPUD, developed at the University of Pennsylvania [Stone, 1998], has the representational and reasoning power needed for generating effective instructions. SPUD s method of generation allows it to flexibly produce descriptions of actions that are linguistically sound, not ad hoc. Using SPUD s ability to reason about the information conveyed by action ....

....about actions in the domain are discussed. These rules for reasoning about actions and agents both contribute to generation of effective instructions. 1. 3 Generating Effective Instructions A Natural Language generator called SPUD (Sentence Planning Using Descriptions) Stone and Doran, 1997; Stone and Webber, 1998; Stone, 1998] developed at the University of Pennsylvania, provides the necessary components for generating effective instructions. SPUD forms descriptions of actions (as well as events, states, and objects) by choosing lexical items from its Lexicalized Tree Adjoining Grammar (i.e. words and ....

[Article contains additional citation context not shown here]

M. Stone and B. Webber. Textual economy through close coupling of syntax and semantics. International Workshop of Natural Language Generation, August 1998.

....termination for Turn the knob include: 90 degrees, to the open position, to open the door, until the door opens. As such information is primarily expressed intra sententially, albeit over multiple clauses, I have begun to use a generation system called SPUD (Stone Doran 1997; Stone Webber 1998) which plans sentences by forming descriptions of events, states, and objects. By matching the semantic and pragmatic information of its Lexicalized Tree Adjoining Grammar entries against information it is told to convey, SPUD chooses those lexical items which serve its communicative goals best in ....

Stone, M., and Webber, B. 1998. Textual economy through close coupling of syntax and semantics. International Workshop of Natural Language Generation.

.... intentions has been used in computational linguistics to show how choices at one level can address multiple goals and allow one to leave some important information unexpressed at another (e.g. clausal connectives at the pragmatic level [DW96] and word choice at the syntax and semantics levels [SW98]) In this paper, we describe our attempts to empirically identify the degree to which intention overloading may affect content selection for object descriptions and what communicative goals might overload with an identification goal. We limit our study to cases in which references are made to a ....

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In workshop7, Niagra-on-the-Lake, Canada, 1998.

....requiring an analysis of the at PP as an NP, which is linguistically unmotivated. 7 Conclusion The goal of our work is to capture lexical semantic properties that we hope will be helpful in reducing the search space in parsing, as well as aid in generation (SPUD; see Stone and Doran 1997; Stone and Webber 1998) and machine translation (in the transfer of lexical semantic properties) see Palmer, et al. to appear) We have examined several subclasses of motion verbs, and posited features to capture their semantic properties. These features not only allow us to place restrictions on the verbs to ....

Stone, M. and B. Webber. (1998). Textual economy through close coupling of syntax and semantics.

No context found.

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In Proceedings of International Natural Language Generation Workshop, pages 178--187, 1998.

.... shared information that the two agents maintain as part of their collaboration, a conversational record [Clark and Marshall, 1981, Thomason, 1990] represented by the operator [CR] The eight formal rules governing these modalities, given in (3) represent a reasonable idealization of conversation [Stone, 1998b] 3) P]p # p [T]p # p [CR]p # p (VER) P]p # [P] P]p [T]p # [T] T]p [CR]p # [CR] CR]p (PI) CR]p # [P]p [CR]p # [T]p (INC) We motivate DIALUP specifications for the patron P and the teller T by considring the DIALUP queries that these agents might use to infer opportunities ....

....of the participants in the conversation. In order to give a more accurate picture of the dynamics of dialogue and the dependencies of utterances on shared context, the logical content of utterances can be represented more precisely, in terms of presuppositions and assertions [van der Sandt, 1992, Stone and Webber, 1998] Moreover, a treatment of dialogues with multiple utterances can be obtained by appealing to AI formalisms of knowledge and ability [Davis, 1994, Stone, 1998a] and introducing a nested implication for each step of action. As outlined in Section 6, these two features allow a much more detailed ....

[Article contains additional citation context not shown here]

Stone, M. and Webber, B. (1998). Textual economy through close coupling of syntax and semantics. In Proceedings of INLG, pages 178--187.

....through the remaining cycles of deliberation and action. The intention representations we arrive at are symbolic, recursive structures that appeal to logical accounts of knowledge and time to characterize actions and their effects in context. Fuller technical details are available in [Stone, 1998, Stone, 2001] But how then can the theory escape the many well known difficulties in using logic to describe the real world The reason is intentions need not provide guarantees about what the world will be like; and they need not represent an agent s predictions about the future. What an ....

....use shared information, the utterance, the grammar, and the attentional and intentional state of the discourse, to reconstruct this interpretation. This is the formulation of the language production problem that I and my colleagues arrived at in the SPUD generation system [Stone and Doran, 1997, Stone and Webber, 1998, Stone et al. 2001] SPUD can generate concise, contextually appropriate utterances, including both speech and concurrent nonverbal behavior, by applying a simple, uniform and efficient decision making strategy. This strategy exploits the lexicalized tree adjoining grammar (LTAG) formalism in ....

Stone, M. and Webber, B. (1998). Textual economy through close coupling of syntax and semantics. In Proceedings of International Natural Language Generation Workshop, pages 178--187.

....information would already be available in an NLIR system s knowledge base (KB) so it is important that it be able to infer this knowledge when given the opportunity. Using a lexicalized grammar, we show that dividing the semantics of lexical items into assertion and presupposition 2 , as in [24] and [26] reveals a simple and elegant analysis which exploits regularities of alt set words. This analysis provides a new approach to interpreting alt set words, including those with discourse anaphora such as (1a) which [8] fails to handle. At the same time, the analysis provides a method for ....

....an utterance to make sense. For an overview of presupposition, see [1] Inference through Alternative Set Semantics 3 (4) I find shoes NP : I (SnNP ) NP : X Y:find(Y; X) NP : shoes SnNP : Y:find(Y; shoes) S : find(I; shoes) 2. 2 Lexical Semantics as Assertion and Presupposition We follow [24] and [26] in separating lexical semantics into assertion and presupposition. The assertion is computed during the derivation (Section 2.1) and presupposition is evaluated through the mechanism described in Section 5.1. We will write lexical entries in the following form, where the semantic ....

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In International Workshop on Natural Language Generation, pages 178-187, Niagaraon -the-Lake, Canada, August 1998.

No context found.

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In Proceedings of INLG, pages 178--187, 1998.

....among them: 3) Remove the bolt with the wrench from the tool box. In (3) one compact and natural sentence describes the goal of an action (removal) its method (wrench) and the knowledge required to accomplish it (the wrench s location) The sentence planning using description (spud) system [27, 28] I describe in section 3.3 also assigns inference about interpretation a central role in generation, albeit a more constrained role than Appelt. These arguments show the advantages that we can expect inference to o#er in ambitious dialogue systems. But what about the kinds of NLG systems that ....

....Sections 3.1 and 3.2 have shown how inferences about knowledge and action can be specified and derived. It remains to be seen how these conclusions might actually be accessed in sentence planning. In this section, I suggest one answer, inspired 1 by the spud system for sentence generation [27, 28]. spud adopts a view of sentence generation as goal directed activity, like [1, 4] before it. On this view, the task of the generator is to use the words and constructions of the language to design a message that fulfills a set of communicative intentions. spud works with two kinds of intentions ....

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In Proceedings of INLG, pages 178--187, 1998.

....ENTRY MUST MAKE IT AS EASY AS POS SIBLE FOR THE GENERATOR TO EXPLOIT ITS CONTRIBUTION IN CARRYING OUT FURTHER PLANNING. This principle responds to two concerns. First, our research has revealed many characteristic uses of language in which a single entry helps achieve multiple communicative goals (Stone Webber, 1998). This is an important way in which a generator needs to be able exploit the contribution of an entry it has already used, in line with our principle. Second, SPUD is currently constrained to greedy or incremental search for reasons of efficiency. At each step, SPUD picks the entry whose ....

....the relationships between sentence meaning, background knowledge and inference relationships which are easiest to state in terms of constraints. In addition, the use of constraints harmonizes with our perspective that a basic generation task is to construct extended descriptions of individuals (Stone Webber, 1998; Webber et al. 1999) In general, to express the semantic links between multiple entries in a derivation, we associate each node in a syntactic tree with the individuals that the node describes. We refer to the collection of individuals that label the nodes in an entry as the SEMANTIC ARGUMENTS ....

[Article contains additional citation context not shown here]

STONE M. &WEBBER B. (1998). Textual economy through close coupling of syntax and semantics. In INLG, p. 178--187.

....(Sequential enablement) 11 7 (Arbitrary causation) 14(a) 15 8 (Non causation) 8(a) 9(a) 15(a) Figure 2.4: Constructions appropriate in each world model 2.3. 3 Generation Generation is done using the SPUD (Sentence Planning Using Descriptions) Natural Language generator [Stone and Doran, 1997; Stone and Webber, 1998] which is described in detail in Section 5.2. SPUD forms descriptions of actions, events, states, and objects, by choosing lexical items from its Lexicalized Tree Adjoining Grammar which serve its communicative goals best. By virtue of being a lexicalized grammar, lexical items correspond to ....

....However, actions can have a linguistically oriented representation associated with them. In fact, there are conceptlexeme mapping structures in the lexicon. Action schemas are used for building a plan, which then can be used as input to the generator. SPUD [Stone and Doran, 1997; Stone and Webber, 1998] focuses on generating contextually appropriate descriptions of entities, much like [Dale, 1992] However, it extends beyond Dale s work since it considers information contributed by the whole sentence to a referring expression. Descriptions are not limited to objects but can be generated for ....

[Article contains additional citation context not shown here]

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. International Workshop of Natural Language Generation, August 1998.

No context found.

Matthew Stone and Bonnie Webber. Textual economy through close coupling of syntax and semantics. In Proceedings of 1998.

No context found.

Stone, M., & Webber, B. (1998). Textual economy through close coupling of syntax and semantics. In Proceedings of INLG (pp. 178--187).

No context found.

Stone, M., and B. Webber: 1998. Textual Economy through Close Coupling of Syntax and Semantics. Proceedings of the 9th INLG Workshop, pp. 178-187.

No context found.

Stone, M. & B. Webber (1998), Textual Economy Through Close Coupling of Syntax and Semantics, Proceedings IWNLG, Montreal, 178-187.

No context found.

Stone, M. & B. Webber (1998), Textual Economy Through Close Coupling of Syntax and Semantics, Proceedings IWNLG, Montreal, 178-187.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC