this technical note and demo we present a new off-theshelf realizer, REALPRO. REALPRO is derived from previous systems (Iordanskja et al., 1988; Iordansk ja et al., 1992; Rambow and Korelsky, 1992), but represents a new design and a completely new implementation. REALPRO has the following characteristics, which we believe are unique in this combination: REALPRO is implemented in C++. It is therefore both fast and portable cross-platform

... document structure can be seen as an extension of Nunberg's `text-grammar'; it is also closely related to `logical' mark-up in languages like HTML and LATEX. We show that by using this intermediate representation, several subtasks in language generation and language understanding can be defined more cleanly.

This report presents a summary of the architectural characteristics of some of the Natural Language Generation (NLG) systems that serve as the main building blocks of the RAGS

What are the benefits of using Natural [anguagc Generation ill an industrial application? We have attempt to answer part of tiffs question with a description of an assessment of three techniques producing multiscntential text: semiautomatic fill-in-lhc-blank interfacing, automalic linguistic-and-templates hybrid generation, and human writing, This assCSSlllClll used a black box methodology, with an indq, ondent blindlosled jury lhal gave cliflrcnt quality levels in rolalion to a set of criteria. The lexis tlsod for /he assessment wctc imsincss reply loltors.

This paper discusses some issues relating to the task of lexicalisation within applied NLG systems. We restrict ourselves here to the systems covered in "A Survey of Applied Natural Language Systems" [Pai98], that are fully implemented, complete

Giving an adequate general definition of the input to natural language generation (NLG), and hence to NLG itself, is a notoriously difficult problem, practically, theoretically and even methodologically. In this paper, we describe our recent experiences of implementing an NLG component of a larger question-answering system, and trying to understand and resolve some of these problems in this context. We examine

The division of text into visual segments such as sentences, paragraphs and sections achieves many

This paper describes SimpleNLG, a realisation engine for English which aims to provide simple and robust interfaces to generate syntactic structures and linearise them. The library is also flexible in allowing the use of mixed (canned and noncanned) representations. 1

Introduction 2 2 L'auto-reponse 3 3 La gestion de courrier electronique 5 4 Les systemes de reponse au texte libre 6 4.1 Systemescommerciaux............................ 6 4.1.1 Answer Agent deBrightware................... 7 4.1.2 Intelliserve deNovator ...................... 11 4.1.3 EchoMail deEchoMailCorporation................ 15 4.1.4 Kana Classify deKana...................... 16 4.1.5 ExpressResponse deIslandData ................. 18 4.1.6 Email Specialist deYYSoftware................. 20 4.1.7 SelectResponse deAptex..................... 25 4.1.8 k-Commerce E-mail deInference................. 26 4.2 Systemes de Question-Reponse........................ 26 4.2.1 AskJeeves............................... 26 4.2.2 Agentsvirtuels-AskEllenetAskRed............... 27 4.2.3 Systemes de recherche en question-reponse............. 29 4.3 Systemes de Generation ..........

Introduction La grande quantit d'information disponible, que ce soit sur Internet ou Intranet, rend ncessaire l'utilisation d'outils informatiques de rsum et de filtrage. Les moteurs de recherche servent chercher l'information : cependant, mme avec ces logiciels, la quantit d'information retrouve est souvent trop grande et les documents proposs sont dans certains cas (encore trop nombreux) peu ou pas du tout pertinents par rapport la recherche effectue. En d'autres termes, on constate le phnomne du bruit. Des outils complmentaires de filtrage sont donc ncessaires. La prsentation d'un rsum orient par la question est aussi de grande utilit pour permettre l'utilisateur de vrifier d'un coup d'oeil si le document trouv est effectivement pertinent pour sa recherche ou pas. Par ailleurs, techniquement, le processus mme de produire un rsum orient par une question est trs li l'estimation fine de la pertinence du document par rapport la question et donc au filtrage de qualit. Dans cette contr