Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Technical Report CUCS-034-94, Columbia University.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for particular needs, the perceived complexity of the user interface, etc. Typical metrics include time and cost to task completion, quality of solution, and user satisfaction, as well as extensibility and portability (e.g. based on coverage of a test set with or without adding new knowledge [32]) It is also useful to carry out regression testing to determine stability of performance of successive system versions on unseen test sets. In the special case of summarization systems, one could also examine whether particular menu items in the user interface are appropriate, whether the ....

Robin, J. 1994. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design and implementation. Ph.D. Thesis, ColumbiaUniversity.

....that satisfy the required lexical constraints. This prevents the lexical chooser from trying to combine incompatible clauses later. By determining sentence boundaries before carrying out detailed lexical decisions, CASPER cuts down the search space of the lexical chooser drastically. In STREAK [Robin, 1995], a generation system which also implements hypotactic aggregation, detailed lexical decisions are made whenever a proposition is aggregated. This is costly because the best lexical decisions. for n propositions might not be useful or correct for n b propositions. The strategy generates ....

.... him . 6 Related Work Both hypotactic and paratactic constructions described in this paper have received a lot of attention in linguistics [Quirk et al. 1985, Halliday, 1994, Carpenter, 1998] Much generation literature on aggregation was disguised under the topic revision [Meteer, 1991, Robin, 1995] Callaway and Lester, 1997] We consider clause aggregation as an integral part of a text gen eration system, not as a revision. The term revision implies that something has been generated and then improved upon, which is not the case in these systems. We prefer the term optimization used by ....

[Article contains additional citation context not shown here]

Robin, J. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Columbia University.

....variant recognition and paraphrase has been done in the fields of natural language generation, text summarization, and information retrieval. The generation community has focused mainly on rule based text transformations in order to meet external constraints such as length and readability [11] 18][22]. Dras [4] described syntactic Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on ....

Robin, J. 1994. Revision-based Generation of Natural Language Summaries Providing Historical Background. Ph.D. Dissertation. Columbia University.

....to each object will be generated as well and used as labels on the illustrated object in the accompanying graphics. Thus, for example, speech just says Two IVs while the textual labels spell out that they are peripheral. MAGIC uses Columbia s FUF SURGE language generation tools for this task [8, 9, 10]. RELATED WORK A key feature of MAGIC is that both content and form of the presentation are dynamically generated at runtime, thus allowing output to be customized for the current user and situation. In contrast, many interfaces for healthcare applications rely on the more traditional approach ....

J. Robin. Revision-Based Generation of Natural Language Summaries: Corpus-Based Analysis, Design, Implementation and Evaluation. PhD thesis, Computer Science Dept., Columbia, 1994.

....description of the text. The linerizer interprets the syntactic description and generates the English sentence. The unification grammar is called SURGE (Systemic Unification Realization Grammar of English) and was developed by the natural language processing group at Columbia University, 13] [14]. It is capable of generating a large variety of sentences. The locative expression generator output is a vector of 0 s, 1 s, and X s, for each of the following prepositions and features: small, medium, big, near, far, above, below, aligned, next, inside, left, right, between) For example, the ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation and Evaluation. PhD thesis, Columbia University, 1994.

....words are selected to realize the values of the fields. The resulting case frame is passed through syntactic generation, where a full syntactic tree of the sentence is created, grammatical constraints are enforced, and morphological agreement is carried out. We use Columbia s FUF SURGE package [5, 16], a robust grammar of English (SURGE) along with a unification interpreter (FUF) and some text manipulation tools written in Perl as the basis for both word choice and syntactic generation. 6 Conclusions and Status In the CDNS project, we have developed tools and a system architecture with which ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis,

.... generation, creating scientific or instructional text which significantly differs in the distribution and frequency of syntactic, semantic, and orthographic features from that found in narrative prose (although a few projects do address some of these issues, e.g. Kantrowitz and Bates, 1992; Robin, 1994; Doran, 1998; Cassell et al. 2000 ] In addition, the most advanced of these systems are still not capable of producing more than two paragraphs of text, while the vast majority of naturally occurring narratives are at least several pages long. Finally, none of these systems are intended to ....

....each FD while maintaining the paragraph separations imposed by the narrative segmenter and proceeds to send paragraph sized batches to the revision component (described in the following section) in order to improve overall prose quality. 3. 3 Revision Revision modules [Dalianis and Hovy, 1993; Robin, 1994; Callaway and Lester, 1997; Shaw, 1998 ] take a series of protosentences (simple sentences with limited content, e.g. The wolf saw Little Red Riding Hood ) and rearrange them by aggregation, i.e. combining protosentences in various ways, or by migration, i.e. permuting the order of two adjacent ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Columbia University, December 1994.

....system previously: Neil Tipper and Rodger Kibble. We are grateful to our anonymous referees for their helpful comments. RAGS, as described in that paper, says very little about the functional structure of an NLG system, or the issues arising from more complex processing regimes (see for example Robin (1994), Inuie et al. 1992) for further discussion) NLG systems, especially end to end, applied NLG systems, have many functionalities in common. Reiter (1994) proposed an analysis of such systems in terms of a simple three stage pipeline. More recently Cahill et al. (1999) attempted to repeat the ....

J. Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background:Corpus-Based Analysis, Design, Implementation and Evaluation. Technical Report CUCS-034-94, Columbia University.

....strongly differentiated, coherence being about communicative intentions and cohesion being about semantic con 1 There are two additional goals that we found in the literature which don t quite fit in with the four groups. They are: a) to make texts convey additional information ( RM93, RM96, Rob94] and, b) to make texts more expressive ( Met90] MKS94] nectedness . We take it that [MB94] is representative of the usage of cohesive in the literature and the community. Cohesion, to Halliday and Hasan, is the defining property of a text qua text what it is that makes a text a ....

Jacques Robin. Revision-based Generation of Natural Language Summaries Providing Historical Background: Corpus-based Analysis, Design, Implementation and Evaluation. Unpublished PhD thesis, Computer Science Department, Columbia University, 1994.

....preserves fluency Simply stringing the phrases together can produce ungrammatical results because phrases are placed in new syntactic contexts. We have developed an algorithm that maps the predicate argument structure from input document phrases to arguments expected by FUF SURGE (Elhadad 1993; Robin 1994), a robust language generation system. This has required developing new techniques to identify constraints on realization choice (e.g. on the order of circumstantial roles such as time, location, instrument, etc. using surface features in place of the semantic or pragmatic ones typically used ....

....combined into a single, more complex sentence, looking again at constraints from the input document as well as common references between phrases. Finally, the constituent structure produced by these two stages is mapped to the functional representation required as input by FUF SURGE (Elhadad 1993; Robin 1994). Document Analysis The first stage of our system identifies similar text units; that is, those that contain important common information. Our definition of similarity is di#erent than the one adopted in most text matching tasks (such as information retrieval) because of two factors: First, the ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. PhD thesis, Department of Computer Science, Columbia University, New York, 1994. Also Columbia University Technical Report CU-CS-034-94.

....in novel contexts Simply stringing the phrases together can produce ungrammatical results because phrases are placed in new syntactic contexts. We have developed an algorithm that maps the predicate argument structure of input document phrases to arguments expected by FUF SURGE [Elhadad 1993; Robin 1994] a robust language generation system. This has required developing new techniques for identifying constraints on realization choice (e.g. on the order of circumstantial roles such as time, location, instrument, etc. using surface features in place of the semantic or pragmatic ones typically ....

....combined into a single, more complex sentence, looking again at constraints from the input document as well as common references between phrases. Finally, the constituent structure produced by these two stages is mapped to the functional representation required as input by FUF SURGE [Elhadad 1993; Robin 1994] . Document Analysis Our definition of similarity is different than the one adopted in most text matching tasks (such as information retrieval) because of two factors: first, the size of the unit of text affects what is similar; documents have a lot of information, so even a modest amount of ....

[Article contains additional citation context not shown here]

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. PhD thesis, Department of Computer Science, Columbia University, New York, 1994. Also Columbia University Technical Report CU-CS-03494.

....we need to nd the di erences from one article to the next, to identify how the reported facts have changed. Thus, one of our main problems was to identify summarization strategies which indicate how information is linked together to form a concise and cohesive summary. As was found in other work [Robin, 1994], what information is included is often dependent on the language available to make concise additions. Thus, using a corpus of summaries was critical to identifying the di erent possible summaries. We have developed a set of planning operators derived from the corpora that we analyzed (Chapter ....

....that case, we need to de ne a concept related to language reuse, namely language regeneration. In the cases when reuse is not sucient, the system has to transform the source text to achieve its communicative goal. Some examples of transformations include sentence simpli cation and nominalization [Robin, 1994]. 80 Ten people were killed overnight by suspected Islamic extremists in the Ain Defla region some 160 kilometers (100 miles) west of the Algerian capital, security services announced Sunday. Shortly before 11:30 a.m. the Dow Jones industrial average was up 85.47 points, or 1 percent, ....

[Article contains additional citation context not shown here]

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, 176 Implementation, and Evaluation. PhD thesis, Department of Computer Science, Columbia University, New York, 1994.

....case role values filled by POS tagged word strings. The draft is then submitted to an automatic text planner which outputs an hierarchical structure of templates which is ordered according to rhetorical and stylistic requirements. This process resembles revision oriented generation (Meteer, 1991, Robin, 1994, Gabriel, 1988, Inui et al. 1992) Using the set of distinctions by Robin, our approach is content preserving (no extra content is added) and performs revisions on a shallow representation. The realization stage linearizes the plan and takes care of the ellipsis, conjoined structures , ....

Robin, J. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background.

....into words, phrases or sentences. It makes extensive use of lexical constraints and grammar. Therefore it is language dependent. Functional Unification [ Kay, 1984 ] is one of the widely used approaches for designing the realization component, such as in the FUF SURGE NLG package [ Elhadad, 1993; Robin, 1994 ] In Functional Unification, the features from two Functional Descriptions (FDs) i.e. nested attributevalue pair structure, are merged using the unification operator. As a result, linguistic constraints at 31 32 APPENDIX A. RELATED WORK di#erent levels, such as discourse, semantic, syntactic, ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Columbia University, 1994.

....are also important. Of course, generality would be of little interest if the overall quality of the text was low, so assessing text quality is critical. We therefore sought to evaluate Fare s text quality as well as it s domain generality and task generality. With four significant exceptions [15, 4, 34, 41], the field of natural language generation has not witnessed the development of an empiricist evaluation school. However, because the conclusions of empirical studies should be considerably less equivocal than those derived from proof of concept systems, we chose the empirical route and ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-based Analysis, Design, Implementation and Evaluation. PhD thesis, Columbia University, September 1994.

....algorithm can achieve 67 correctness simply by guessing the every time. On the other hand, human experts cannot achieve much better than 95 faced with the same task. This puts Knight and Chander s achievement of 78 into perspective. 4.5. How to Get Adequate Training and Test Data As Robin (1994) points out, such is the complexity of NLG that many people have focussed on particular subtasks which have been evaluated using hand coded (and hence few in number) inputs. Few projects have had the benefit of a significant knowledge base of the kind used by Acker, Lester and Porter from which to ....

Robin, J. (1994) "Revision-Based Generation of Natural LanguageSummaries Providing Historical Background: CorpusBased Analysis, Design, Implementation and Evaluation", PhD thesis, Columbia University, 1994; also available as Technical Report CUCS-034-94.

....we need to find the di#erences from one article to the next, to identify how the reported facts have changed. Thus, one of our main problems was to identify summarization strategies which indicate how information is linked together to form a concise and cohesive summary. As was found in other work [Robin, 1994], what information is included is often dependent on the language available to make concise additions. Thus, using a corpus of summaries was critical to identifying the di#erent possible summaries. We have developed a set of planning operators derived from the corpora that we analyzed (Chapter ....

....case, we need to define a concept related to language reuse, namely language regeneration. In the cases when reuse is not su#cient, the system has to transform the source text to achieve its communicative goal. Some examples of transformations include sentence simplification and nominalization [Robin, 1994]. 80 Ten people were killed overnight by suspected Islamic extremists in the Ain Defla region some 160 kilometers (100 miles) west of the Algerian capital, security services announced Sunday. Shortly before 11:30 a.m. the Dow Jones industrial average was up 85.47 points, or 1 percent, ....

[Article contains additional citation context not shown here]

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, 176 Implementation, and Evaluation. PhD thesis, Department of Computer Science, Columbia University, New York, 1994.

....information. FUF is an extension of FUG and includes more control features and type hierarchy motivated by the speci c needs of generation: lexicalization, phrase planning and even non linguistic purposes that have appeared during the development of practical systems. According to Robin [11], the bene ts of using FUF for developing generation systems (as opposed to a direct syntactic encoding of the linguistic information) are: 1. The formalism supports partial information, is declarative, uniform and compact. 4 CHAPTER 2. BACKGROUND 5 2. The same formalism can be used for all ....

....FUF SURGE [4] 5] and the FUF SURGE package is used in many projects in NLG research. Using the FUF SURGE package, implementing a generation system consists of decomposing non syntactic processing into sub processes and encoding in FUF the knowledge sources for each of these sub tasks (see [11] for a complete discussion) From a linguistic perspective, the functional analysis of language (in contrast to a structural analysis) identi es the function of each substring withing the larger string of words. The structural stage of this analysis identifyies the constituents of a phrase. For ....

J. Robin. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. Technical Report CU-CS-034-94, Computer Science Department, Columbia Universtity, New York, NY, 1994. PhD. Thesis.

....and semantic knowledge in order to produce meaningful and uent output. Such knowledge is often hand coded anew when a di erent application is developed. We present in this paper the integration of a large scale, reusable lexicon with a natural language generator, FUF SURGE (Elhadad, 1992; Robin, 1994); we show that by integrating the lexicon with FUF SURGE as a tactical component, we can reuse the knowledge encoded in the lexicon and automate to some extent the development of the lexical realization component in a generation application. The integration of the lexicon with FUF SURGE also ....

....1 [2] 2 [3] 3 [4] # 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 Figure 5: Mapping subcategorization NP WITHNP to SURGE input of uni cation. 5 Related Work The lexicon, after it is integrated with FUF SURGE, can also be used for other tasks in language generation. For example, revision (Robin, 1994) is a technique for building semantic inputs incrementally. The revision process decides whether it is appropriate to attach a new constituent to the current semantic input, for example, by adding an 214 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 relation concept become visible word appear a ....

J. Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. Ph.D. thesis, Department of Computer Science, Columbia University.

....the text. The linerizer interprets the syntactic description and generates the English sentence. The unification grammar is called SURGE(Systemic Unification Realization Grammar of English) and was developed by the natural language processing group at Columbia University, McKeown et al. 1990] [Robin, 1994]. It is capable of generating a large variety of sentences. Before the natural language generator can generate the English sentence the language generation preprocessor must translate the output of the locative expression generator. The locative expression generator s output is a vector of 0 s, ....

Jacques Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation and Evaluation. PhD thesis, Columbia University, 1994.

....clauses( sub( Although the rules in (96) are sufficient for producing text that exercises a broad range of intonational possibilities, other generation approaches have incorporated more detailed rules that might reasonably replace (96)c and (96)d. Robin (1993, 1994), and Robin and McKeown (1993) employ revision based techniques that allow additional information to be added at the sentence planning level. Such Additional Deep Syntactic Specifications (ADSSs) are similar to the embedded properties in Example 6.2. Robin s approach allows these embedded ....

....(101) which constructs appropriate referring expressions. Other researchers (Reiter 1991; Smadja and McKeown 1991; Elhadad and Robin 1992; Elhadad, McKeown and Robin 1996) have studied the problem of lexical choice in generation frameworks with encouraging results. For example, the STREAK system (Robin 1994) is able to produce the sentences in (103)a and (103)b, depending on whether the manner constraint is to be conveyed by the prepositional phrase or the verb. Such techniques do not conflict with the information structural aspects of the present sentence planner, and therefore stand to be ....

Robin, J. (1994). Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation and Evaluation.

....is the MUC 6 management transition domain. What are paraphrases Literally speaking, paraphrases are the different ways of expressing the same fact, or event. Paraphrases can be of arbitrary complexity. The major work of using paraphrases in natural language processing is for language generation [93] and text summarization [35] The paraphrases are applied to generate summaries and improve texts. Examples are the passive to active voice paraphrase, relative pronoun deletion and avoidance of nominalization, etc. In the context of text summarization, a paraphrase can be defined as consisting of ....

Jacques Robin. Revision-based generation of natural language summaries providing historical background. Ph.D Disseration, 1994.

....to handle this sort of paraphrase. There are existing Natural Language Processing systems which already deal with surface constraints on text. Most of these are natural language generation (NLG) systems, and fall within the field of revision based generation: weiveR (Inui et al., 1992) and Streak (Robin, 1994) are two such systems. In these systems, draft texts are produced, criticised, and then paraphrased 1 to produce a text that fits certain constraints. However, in these systems the constraints are local to a small unit of text, generally a sentence. For example, Streak has a constraint that ....

....for academic conferences, and meeting this constraint often involves cutting down a longer draft version. Constraining text length is also a feature of computational language generation systems, for instance as an explicit limit on the length of an individual text unit, as in the Streak system (Robin, 1994). Another common measure comes from readability formulae, such as the Flesch Reading Ease Score (Klare, 1974 5) Standard readability formulae are equations which attempt to predict, rather than evaluate, the readability of text; in form they are generally linear combinations of factors which ....

Robin, J. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background.

....following grammatical structures: Noun Group (NP) Post Positional group (PP) Adverb Group (AdvG) More information about these units can be found in Sections 3.5, 3.6, and 3.7, respectively. The disjuncts and conjuncts are presented in [38] and most of them are implemented in SURGE 2. 0 (see [9, 17]) Circumstantial functions can be decomposed into seven classes (as shown in Table 3.12) according to their functional features. The number of these classes may increase by depending on linguistic analysis. Class Semantic Func. Answer the Ques. Realizations Spatial direction in what direction ....

R. Jacques. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. PhD thesis, Department of Computer Science, Columbia University, New York, USA, 1994.

....we intend to pursue is the rescoring of the top N generated sentences by more expensive (and extensive) methods, incorporating for example stylistic features or explicit knowledge of flexible collocations. Acknowledgments We would like to thank Yolanda Gil, Eduard Hovy, Kathleen McKeown, Jacques Robin, Bill Swartout, and the ACL reviewers for helpful comments on earlier versions of this paper. This work was supported in part by the Advanced Research Projects Agency (Order 8073, Contract MDA904 91 C 5224) and by the Department of Defense. ....

Jacques Robin. 1995. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-based Analysis, Design, Implementation, and Evaluation. Ph.D. thesis, Computer Science Department, Columbia University, New York, NY. Also, Technical Report CU-CS034 -94.

....this success, empirical results in explanation generation are limited. Although techniques for developing and evaluating robust explanation generation should yield results that are more conclusive than those produced by prototype, proof of concept systems, with only a few notable exceptions [27, 23, 8, 46, 54], most work has adopted a research methodology in which a proof of concept system is constructed and its operation is analyzed on a few examples. While isolating one or a small number of problems enables researchers to consider particular issues in detail, it is difficult to gauge the scalability ....

....discourse knowledge representation. We discuss each of these in turn. Evaluation Methodologies. With regard to evaluation, Knight is perhaps most closely related to five NLG projects that have been empirically evaluated: Pauline [23] Edge [8] the Example Generator 17 [46] Ana [27] and Streak [54]. By varying pragmatic information such as tone, Hovy enabled Pauline to generate many different paragraphs on the same topic. Pauline s texts were not formally analyzed by a panel of judges, and it did not produce texts on a wide range of topics (it generated texts on only three different ....

[Article contains additional citation context not shown here]

J. Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Columbia University, December 1994.

....No. N00014 94 1 0338 to Illinois Institute of Technology. The content does not reflect the position or policy of the government and no official endorsement should be inferred. Reva Freedman is now at the Learning Research and Development Center of the University of Pittsburgh. suggested by Robin (1994) and others, the turn planner converts the tutorial planner primitives to one or more semantic forms representing the concepts the tutor wants to convey. These semantic forms are then realized as surface text. For further details about the CIRCSIM Tutor planner see Freedman, 1996. The basic ....

Robin, J. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-based Analysis, Design, Implementation and Evaluation. Ph.D. diss., Columbia University.

....that satisfy the required lexical constraints. This prevents the lexical chooser from trying to combine incompatible clauses later. By determining sentence boundaries before carrying out detailed lexical decisions, Casper cuts down the search space of the lexical chooser drastically. In STREAK [Robin, 1995], a generation system which also implements hypotactic aggregation, detailed lexical decisions are made whenever a proposition is aggregated. This is costly because the best lexical decisions for n propositions might not be useful or correct for n 1 propositions. The strategy generates ....

.... him . 6 Related Work Both hypotactic and paratactic constructions described in this paper have received a lot of attention in linguistics [Quirk et al. 1985, Halliday, 1994, Carpenter, 1998] Much generation literature on aggregation was disguised under the topic revision [Meteer, 1991, Robin, 1995] Callaway and Lester, 1997] We consider clause aggregation as an integral part of a text generation system, not as a revision. The term revision implies that something has been generated and then improved upon, which is not the case in these systems. We prefer the term optimization used by ....

[Article contains additional citation context not shown here]

Robin, J. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Columbia University.

....generation are limited. Although techniques for developing and evaluating robust explanation generation should yield results that are more conclusive than those produced by prototype, proof of concept systems, with only a few notable exceptions (Kukich 1983; Hovy 1990; Cawsey 1992; Mittal 1993; Robin 1994), most work has adopted a research methodology in which a proof of concept system is constructed and its operation is analyzed on a few examples. While isolating one or a small number of problems enables researchers to consider particular issues in detail, it is difficult to gauge the scalability ....

....its most important contribution lies in its evaluation methodology. With regard to evaluation, it is perhaps most closely related to five NLG projects that have been empirically evaluated: Pauline (Hovy 1990) Edge (Cawsey 1992) the Example Generator (Mittal 1993) Ana (Kukich 1983) and Streak (Robin 1994). Pauline s texts were not formally analyzed by a panel of judges, and it did not produce texts on a wide range of topics (it generated texts on only three different events. nevertheless, it is a significant achievement in terms of evaluation scale because of the sheer number of texts it ....

Robin, J. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background.

....using modifiers such as adjectives, and paratactic operators which create, for example, conjunctions) is enabled through lookahead to the lexicon used during realization. The speech generator also includes a realization component, implemented using the FUF SURGE sentence generator (Elhadad, 1992; Robin, 1994), which produces the actual language to be spoken as well as textual descriptions that are used as labels in the visual presentation. It performs lexical choice and syntactic realization. Our version of the FUF SURGE sentence generator produces sentences annotated with prosodic information and ....

....(Horacek, 1992) aggregation is performed at text structure level. In addition to conjoining VP and NPs, FlowDoc(Passonneau et al. 1996) uses ontological generalization to combine descriptions of a set of objects into a more general description. Based on a corpus analysis in the basketball domain, (Robin, 1994) catalogued a set of revision operators such as adjoin and nominalization in his system STREAK. Unlike STREAK, MAGIC does not use revision to combine information in a sentence. Generating spoken language from meanings or concepts (Meaning to Speech, MTS) is a new topic and only a few such systems ....

J. Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Ph.D. thesis, Computer Science Department, Columbia University.

.... the high level sentence structure of each sentence and the words which realize each semantic role, and ffl the FUF (Functional Unification Formalism) Elhadad 1991; Elhadad 1993] sentence generator, which uses a large systemic grammar of English, called SURGE 3 [Halliday 1985; Elhadad 1993; Robin 1994] to fill in syntactic constraints, build a syntactic tree, choose closed class words, and eventually linearize the tree as a sentence. Input to SUMMONS is a set of templates, where each template represents the information extracted from one or more articles by a message understanding system. We ....

.... system, developed jointly with Bellcore to summarize the activities of telephone planning engineers [McKeown et al. 1994] In particular, we used FUF to implement the lexical chooser, representing the lexicon as a grammar as we have done in many previous systems (e.g. Elhadad 1993; Robin 1994; McKeown et al. 1993; Feiner and McKeown 1991] and thus the main effort was in identifying the words and phrases needed for the domain. The content planner, implemented in PERL, features several stages, as does the PLANDoc system. It first groups messages together, identifies commonalities ....

[Article contains additional citation context not shown here]

J. Robin. Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD thesis, Computer Science Department, Columbia University, 1994.

No context found.

Robin, J. 1994b. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. PhD. Thesis. Available as Technical Report CU-CS-034-94. Computer Science Department, Columbia University, New York.

....the state of language generation. In this paper, we first present our new revision based model for summary generation, describing the corpus analyses that led to it, then we move to the evaluation of our approach. We focus on two aspects of our research which is presented in full detail in [47]: 1) its empirical flavor and (2) the quantitative evaluation of same domain robustness. Other aspects of the work are discussed in other publications. For example, see [45] for details on the novel approach to sentence planning (and in particular its motivation and comparison with previous ....

....of same domain robustness. Other aspects of the work are discussed in other publications. For example, see [45] for details on the novel approach to sentence planning (and in particular its motivation and comparison with previous approaches) 46] for an overview of the full project. See also [47] for a quantitative evaluation of cross domain portability of the revision rules underlying our new generation model. It shows that 31 out of the 57 branches (69.5 ) in the class hierarchy of revision rules that we abstracted from the basketball corpus are also used in stock market reports. Given ....

[Article contains additional citation context not shown here]

J. Robin. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. Technical Report CU-CS-034-94, Computer Science Department, Columbia Universtity, New York, NY, 1994. PhD. Thesis.

....are mergeable as composite processes, 3) to constrain syntactic alternations and (4) to constrain the syntactic realization of each participant. surge currently covers 21 simple process types and 15 composite process types, thus accepting 36 different nuclear thematic structures as input (cf. [32] for examples of each type) The Mood System The subtasks of the mood system are: 1) to provide an interface to the specification of speech acts and interpersonal constraints in the client program (e.g. imperative mood to express a request to a subordinate) 2) to account for hypotactic ....

....subordination, embedding) and (3) to constrain to use of abbreviated forms (e.g. participle and verbless clauses) The mood of dependent clauses is often inferred by surge from its syntactic function in the matrix and the head verb of the matrix. surge currently covers 15 different moods (cf. [32] for examples of each mood) The Adverbial System The main tasks of the adverbial system are: 1) to provide an interface to the semantic encoding used in the client program, 2) to determine the relative ordering of adverbials within the clause, 3) to restrict possible co occurrences of ....

[Article contains additional citation context not shown here]

J. Robin. Revision-based generation of natural language summaries providing historical background: corpus-based analysis, design, implementation and evaluation. Technical Report CU-CS-034-94, Computer Science Department, Columbia Universtity, New York, NY, 1994. PhD. Thesis.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Technical Report CUCS-034-94, Columbia University.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Ph.D. thesis, Department of Computer Science, Columbia University, New York.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. Ph.D. thesis, Department of Computer Science, Columbia University, New York. Also Columbia University Technical Report CU-CS-034-94.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Ph.D. thesis, Columbia University.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Ph.D. thesis, Computer Science Department, Columbia University.

No context found.

Jacques Robin. RevisionBased Generation of Natural Language Summaries: Corpus-Based Analysis, Design, Implementation and Evaluation. PhD thesis, Computer Science Department, Columbia University, forthcoming.

No context found.

Robin, Jacques. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation and Evaluation. Ph.D. thesis, Department of Computer Science, Columbia University. Technical Report CUCS-034-94.

No context found.

"Revision-Based Generation of Natural Language Summaries Providing Historical Background", Columbia University PhD Thesis, CUCS-034-94, New York

No context found.

"Revision-Based Generation of Natural Language Summaries Providing Historical Background", Columbia University PhD Thesis, CUCS-034-94, New York

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. Ph.D. thesis, Department of Computer Science, Columbia University, New York. Also Technical Report CU-CS034 -94.

No context found.

Jacques Robin [1994] Revision-Based Generation of Natural Language Summaries Providing Historical Background. PhD Thesis, Columbia University.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-Based Analysis, Design, Implementation, and Evaluation. Ph.D.

No context found.

Robin, Jacques. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background: Corpus-based analysis, design, implementation and evaluation. Ph.D. thesis, Computer Science Department, Columbia Univ.

No context found.

Jacques Robin. 1994. Revision-Based Generation of Natural Language Summaries Providing Historical Background. Ph.D. thesis, Computer Science Department, Columbia University.

First 50 documents  Next 50

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