Sibun, P. (1992). Generating text without trees. Computational Intelligence, 8(1):102--122. 167  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... [Mann and Thompson, 1988] What is the relationship between coherence, attention and the selection of referring expressions [Grosz and Sidner, 1986, Grosz et al. 1995] Is discourse structured hierarchically, linearly or otherwise [Hurtig, 1977, Hinds, 1979, Skorochod ko, 1972, Webber, 1991, Sibun, 1992] What are the relationships between the utterances in multi party discourse [Hobbs, 1983, Walker and Whittaker, 1990] How can large scale shifts in narrative texts be identified [Grimes, 1975, Youmans, 1990] Since we cannot review the entire discourse structure literature, we will summarize ....

Sibun, P. (1992). Generating text without trees. Computational Intelligence, 8(1):102--122. 167

....has been addressed us ing focus preferences for selecting from a content pool, these preferences being embodied in a selection mechanism. Other approaches exploited the structure of domain knowledge with mechanisms for traversing data structures representing this knowledge [Paris McKeown 1986, Sibun 1992] Planning approaches initially utilized more local yet still schematic specifications of ordering, expressed as preconditions or optional satel lites for plan operators [Cawsey 1989, Hovy 1988, Moore 1989] More recently, partial order causal link (POCL) planning is being applied to discourse ....

.... temporal ordering inhibits comprehension [Irwin 1980] and disrupts thematic processing [Townsend 1983] It also facilitates the hearer s identification of implicitly expressed temporal relations [Lascarides Oberlander 1992] Related strategies are available for spatial descriptions [Linde 1974, Sibun 1992] Differentla and Context. The differentia relation holds between two propositions when one proposition (P s ) differentiates a subclass s from other subclasses of a class c. An explainer chooses the statement (P s , from amongst all the possible predicates one could apply to s because P ....

P. Sibun. Generating text without trees. Computational Intelligence 8, 1992.

....potential (Figure 5) The use of interest and importanc e in ILEX is. analogous to the use of salience in [McDonald and Conklin 82] Because the process is seen as a graph traversal . problem, there are als0 similarities with work on generating text from semantic networks [Simmons and Slocum 72, Sibun 92] In a sense, our work aims to combine the best of both. 3.3 Text Planning Although the.pracess of content determination has worked through a number of moves that may be made in the generated text, the result is not the kind of tree structure that one needs for realisation and also has bee n ....

Sibun, P., "Generating texts without trees", Computational Intelligence,8, 102-122, 1992.

....other span. Here, adherence to these assumptions results in a well structured text. 3 Some structural problems with elaboration While these assumptions are useful for a text planner, they are also problematic in a number of respects, as has often been noted: see for instance Mooney et al. (1990) Sibun (1992) and Kittredge et al. (1991) Our central concern in this paper is to associate these problems with one relation in particular, namely the RST relation known as (object attribute) elaboration. 1 Mann and Thompson de ne this to hold between two spans if the nucleus presents an object (i.e. ....

....structure extends to registers beyond the descriptive genre we are considering. This question can be considered by comparing the goal structure in our texts with that in other texts. The communicative goal in our texts is to describe a domain of inter related entities. The genre considered in Sibun (1992) is somewhat similar. In such cases, it is perhaps better to think of a set of goals to describe individual entities, rather than a single overriding goal. The relationships between these goals are not exactly hierarchical, as they are in texts whose aims are principally argumentative or ....

Sibun, P. (1992). Generating text without trees. Computational Intelligence, 8(1), 102-122.

....determine the semantic content to be conveyed. Once the content is determined, other processes are typically used to map the semantic representations onto the words the agent actually speaks. Other approaches have included the use of rhetorical relations (Hovy 1988) and domain specific knowledge (Sibun 1992) to guide the production of propositional output, but these are no more adequate to planning for social dialog. Given the novel requirements that social discourse places upon a conversational agent s action selection mechanism, in particular the requirements for pursuing multiple, nondiscrete ....

Sibun, P. (1992). "Generating Text Without Trees." Computational Intelligence: Special Issue on Natural Language Generation 8(1): 102-122.

.... [36, 4, 47] providing advice about degree planning [32] recommending computer program enhancements [35] interacting with users of database systems [31, 16] generating arguments [27] explaining algebraic manipulations [50] responding to misconceptions [28] and describing spatial knowledge [43]. Because text planners will be developed for an increasingly broad set of applications, each of which will produce message 7 Text Planner Message Specification FD Skeleton Retriever FD Skeleton Processor FD Skeleton Library Lexicon Noun Phrase Generator Functional Descriptions ....

Penelope Sibun. Generating text without trees. Computational Intelligence, 8(1):102-- 122, March 1992.

....Sidner, 1986] 30 September Coherence, Intentions and Discourse Structure: RST and Intentions. Basic reading: Moore and Pollack, 1992; Moore and Paris, 1993] Additional reading: Moser and Moore, 1995; Moser and Moore, 1996] 2 October Information and Intention in Text Generation Basic reading: [Sibun, 1994; Young et al. 1994] Additional reading: Knott et al. 1997] Presenter: Julie Bourne 7 October ) 10 October (3:30 5pm) Discourse Models, Context and Reference Resolution (I) Centering Basic reading: Grosz et al. 1995; Brennan et al. 1987] 9 October ) 14 October (4:30 6pm) Discourse ....

Penni Sibun. Generating text without trees. Computational Intelligence, 8(1):102--122, 1994.

....determine the semantic content to be conveyed. Once the content is determined, other processes are typically used to map the semantic representations onto the words the agent actually speaks. Other approaches have included the use of rhetorical relations (Hovy 1988) and domain specific knowledge (Sibun 1992) to guide the production of propositional output, but these are no more adequate to planning for social dialog. Given the novel requirements that social discourse places upon a conversational agent s action selection mechanism, in particular the requirements for pursuing multiple, non discrete ....

Sibun, P. (1992). "Generating Text Without Trees." Computational Intelligence: Special Issue on Natural Language Generation 8(1): 102-122.

....with elaboration As well as being simple and parsimonious, RST s theory of span structure is able to account for the coherence of a large number of texts. However, the theory has been criticised from several perspectives. For instance, the assumption of tree structure has been questioned by Sibun (1992), and the assumption of continuous constituency has been questioned by Kittredge et al. (1991) Our central concern in this paper is to associate these structural problems with one RST relation in particular, namely object attribute elaboration. 1 Mann and Thompson de ne this relation to hold ....

Sibun, P. (1992). Generating text without trees. Computational Intelligence, 8(1), 102-122.

....information. A variation of the schemata approach (McKeown 1985, 1986) ensures adherence to certain domain and genre specific constraints concerning the presentation of information. Within those constraints, the order in which information is conveyed depends on domain specific knowledge (as in Sibun 1991, 1992), the communicative intention of the speaker and beliefs about the hearer s knowledge. Finally, an approach based on rhetorical structure theory is used to rearrange propositions in order to make certain rhetorical relationships among them, such as contrasts, salient. The implementation, which is ....

....imposing syntactic constraints along the way. This incremental realization of hierarchical structures and depth first traversal is reminiscent of earlier work by McDonald (1986) on description directed control. Another novel approach to the high level generation planning problem is offered by Sibun (1991,1992). In her approach, plans for describing architectural floor plans are generated without producing discourse trees. Propositions are linked to one another not by rhetorical relations or pre planned templates, but rather by the structure of the domain, including physical and spatial properties ....

[Article contains additional citation context not shown here]

Sibun, P. (1992). Generating text without trees. Computational Intelligence: Special Issue on Natural Language Generation, 8(1).

.... Content Determination 5) The use of interest and importance in ILEX is analogous to the use of salience in [McDonald and Conklin 82] Because the process is seen as a graph traversal problem, there are also similarities with work on generating text from semantic networks [Simmons and Slocum 72, Sibun 92] In a sense, our work aims to combine the best of both. 3.3 Text Planning Although the process of content determination has worked through a number of moves that may be made in the generated text, the result is not the kind of tree structure that one needs for realisation and also has been ....

Sibun, P., "Generating texts without trees", Computational Intelligence, 8, 102--122, 1992.

....as reflected in discourse cues, in a finer grained pass. The approach advocated here is similar in that it does not make a priori assumptions about how the text is structured. Many texts are best characterized as having a sequential structure; a hierarchical assumptioncould be misleading ((Sibun 1992) discusses this point in the context of generating spoken text) Separating tiling from a subsequent functional role analysis stage might make discerning these roles easier. Givon 1983) presents a quantitative approach to discourse analysis which focuses on identification of syntactic clues, such ....

Sibun, P. (1992). Generating text without trees. Computational Intelligence: Special Issue on Natural Language Generation, 8(1):102--122.

.... get [from place 1 ] to place 2 This last question is answered at present with less than satisfactory language generation that simply explains the rops that form the path; stepping this answer up to a higher level involves performing better rop merging and inference, perhaps along the lines of (Sibun 1992). To make a mobile robot usable by people other than its designers, it must be possible to interact with it in a reasonably intuitive way. It should not be necessary to understand the robot s internal representations in order to train it in a new environment. A significant obstacle to making ....

Sibun, P. 1992. Generating text without trees. Computational Intelligence 8(1):100--102.

....difficulties in linguistically expressing spatial relationships, generating spatial natural language poses enormous difficulties. While foundational work has studied generating spatial natural language, e.g. scene description generation (Novak 1987) and spatial layout description generation (Sibun 1992), the interplay between relative and absolute coordinate systems must be carefully monitored. For example, in explaining how magnets induce a field that flows from the north pole of a magnet to a south pole of another magnet, and explaining how current in a wire flows from positive electrodes to ....

Sibun, P. 1992. Generating text without trees. Computational Intelligence 8(1):102--122.

.... is appropriate for the user s level of expertise [52] and Romper s schemata include information about the content of proposi 19 A third alternative proposed by Sibun are short range strategies that exploit relations such as spatial proximity to guide the generator through the domain knowledge [55]. These strategies are designed for flexibility, but they do not account for extended explanations, which require a more global rhetorical structure. tions to be selected, as well as their communicative role. 20 Although schemata have been criticized because they lack flexibility, they ....

P. Sibun. Generating text without trees. Computational Intelligence, 8(1):102--122, March 1992.

....by the natural language generation and intelligent multimedia communities. Natural language researchers have studied reference generation, e.g. Dale s classic work on referring expressions [Dale, 1992] scene description generation [Novak, 1987] and spatial layout description generation [Sibun, 1992]. Work on intelligent multimedia systems [Andr e et al. 1993, Feiner and McKeown, 1990, Maybury, 1991, Roth et al. 1991, Mittal et al. 1995] has produced techniques for dynamically incorporating highlights, underlines, 1 The same study employed pre and post tests to evaluate learning ....

Sibun, P. (1992). Generating text without trees. Computational Intelligence, 8(1):102--122.

....has been addressed using focus preferences for selecting from a content pool, these preferences being embodied in a selection mechanism. Other approaches exploited the structure of domain knowledge with mechanisms for traversing data structures representing this knowledge [Paris McKeown 1986, Sibun 1992] Planning approaches initially utilized more local yet still schematic specifications of ordering, expressed as preconditions or optional satellites for plan operators [Cawsey 1989, Hovy 1988, Moore 1989] More recently, partial order causal link (POCL) planning is being applied to discourse ....

.... temporal ordering inhibits comprehension [Irwin 1980] and disrupts thematic processing [Townsend 1983] It also facilitates the hearer s identification of implicitly expressed temporal relations [Lascarides Oberlander 1992] Related strategies are available for spatial descriptions [Linde 1974, Sibun 1992] Differentia and Context. The differentia relation holds between two propositions when one proposition (P s ) differentiates a subclass s from other subclasses of a class c: An explainer chooses the statement (P s ) from amongst all the possible predicates one could apply to s because P ....

P. Sibun. Generating text without trees. Computational Intelligence 8, 1992.

....relations (Halliday Hasan 1976) in a manner similar to that suggested by Skorodch ko. 2 Interestingly, Chafe arrived at the Flow Model after working extensively with, and then becoming dissatisfied with, a Longacre style hierarchical model of paragraph structure (Longacre 1979) 3 cf. (Sibun 1992) for a discussion of how the form of people s descriptions often mirror the form of what they are describing. This differs from the work of (Morris Hirst 1991) in several ways, the most important of which is that our algorithms emphasize the interaction of multiple simultaneous themes, rather ....

Sibun, Penelope. 1992. Generating text without trees. Computational Intelligence: Special Issue on Natural Language Generation 8.102--122.

No context found.

Sibun, P. (1992) Generating Text without Trees. Computational Intelligence 8(1) 102-122.

No context found.

Penelope Sibun. Generating text without trees. ComputationalIntelligence, 8(1):102--122, February 1992.

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