Richard I. Kittredge

Using natural-language processing to produce weather forecasts

Improved numerical weather prediction simulations have led weather services to examine how and where human forecasters add value to forecast production. The Forecast Production Assistant (FPA) was developed with that in mind. The authors discuss the Forecast Generator (FOG), the first application developed on the FPA. FOG is a bilingual report generator that produces routine and special purpose forecast directly from the FPAs graphical weather predictions. Using rules and a natural-language generator, FOG converts weather maps into forecast text. The natural-language issues involved are relevant to anyone designing a similar system.<<ETX>>

Lexical Selection and Paraphrase in a Meaning-Text Generation Model

We introduce a computationally tractable model for language generation based on the Meaning-Text Theory of Mel’cuk et al., in which the lexicon plays a central role. To illustrate the descriptive scope and paraphrase capabilities of the model, we show how the lexicon influences the set of choices at four different points during the multi-stage realization process: (1) semantic net simplification, (2) determination of root lexical node for the deep syntactic dependency tree, (3) possible application of deep paraphrase rules using lexical functions, and (4) surface syntactic realization. We also show some of the ways in which the theme/rheme specifications within the semantic net influence lexical and syntactic choices during realization. Examples are taken primarily from an implemented system which generates paragraph-length reports about the usage of operating systems.

On the need for domain communication knowledge

Much research in text planning has been oriented towards identifying the rhetorical component of communication knowledge that allows for the structuring of coherent multi‐sentential or even multi‐paragraph text. However, several researchers have recently found rhetoric‐based top‐down approaches to be inadequate for their particular needs. In this paper, we explore the issue by introducing the concept of “domain communication knowledge” as a new analytical tool. By analyzing two existing rhetoric‐based text planners, and by closely analyzing a particular genre, namely reports, we conclude that domain‐specific knowledge about communication is pervasive and should be represented explicitly for the text planning task.

Bilingual generation of weather forecasts in an operations environment

In 1986 the first experiments in text generation applied to weather forecasts resulted in a prototype system (t~AREAS[6,3]) for producing English marine bulletins from forecast data. Subsequent work in 1987 added French output to make the initial system bilingual (RAREAS-2111]). During 1988-1989 a full-scale operational system was created to meet the needs of daily marine forecast production for three regional centres in the Canadian Atmospheric Environment Service 1. In contrast to the earlier systems, the most recent one uses general models for both text planning and sentence realization (see sections 4 and 5 below).

Handling Stuctural Divergences and Recovering Dropped Arguments in a Korean/English Machine Translation System

This paper describes an approach for handling structural divergences and recovering dropped arguments in an implemented Korean to English machine translation system. The approach relies on canonical predicate-argument structures (or dependency structures), which provide a suitable pivot representation for the handling of structural divergences and the recovery of dropped arguments. It can also be converted to and from the interface representations of many off-the-shelf parsers and generators.

Synthesizing weather forecasts from formated data

This paper describes a system (RAREAS) which synthesizes marine weather forecasts directly from formatted weather data. Such synthesis appears feasible in certain natural sublanguages with stereotyped text structure. RAREAS draws on several kinds of linguistic and non-linguistic knowledge and mirrors a forecasters apparent tendency to ascribe less precise temporal adverbs to more remote meterological events. The approach can easily be adapted to synthesize bilingual or multi-lingual texts.

Semantic processing of texts in restricted sublanguages

Abstract Practical results in information retrieval and automatic translation have recently been achieved for naturally-occurring texts in certain narrow technical areas. For each application, the processing system must exploit the distinctive linguistic properties of the appropriate sublanguage; in fact, a precise description of these properties, incorporated into a sublanguage grammer and lexicon, is what enables the system to build a representation of the information (meaning) conveyed by the text. Sublanguages which appear insufficiently closed for semantic processing often carry an important component of information which is encoded in a linguistically well-behaved way and is hence computationally separable. By way of illustration, a procedure is outlined for processing stock market reports into a predicate-argument representation of their content, for that part of the report which refers to the stock exchange activity. The procedure may have applications beyond information retrieval, in particular to the synthesis of informative stock market reports in one or more languages.

Generating Grammatical and Lexical Anaphora in Assembly Instructional Texts

In this paper, we discuss the problem of generating natural anaphora in assembly instructional texts. We first present a detailed account of grammatical and lexical anaphora and we examine a set of constraints for selecting these devices. As language is often redundant and an optimal referring expression is computationally prohibitive, we take the view that the generation of anaphora should be based on a thorough linguistic study and should lead to a natural choice, rather than an optimal one. We then present a text generation system built to evaluate the linguistic constraints. This component takes as input a specification of the sequence of actions to be performed in a cooking recipe and selects and produces the most appropriate anaphoric device to be used in the text.

SEMANTIC PROCESSING OF TEXTS IN RESTRICTED SUBLANGUAGES

Practical results in information retrieval and automatic translation have recently been achieved for naturally-occurring texts in certain narrow technical areas. For each application, the processing system must exploit the distinctive linguistic properties of the appropriate sublanguage; in fact, a precise description of these properties, incorporated into a sublanguage grammer and lexicon, is what enables the system to build a representation of the information (meaning) conveyed by the text. Sublanguages which appear insufficiently closed for semantic processing often carry an important component of information which is encoded in a linguistically well-behaved way and is hence computationally separable. By way of illustration, a procedure is outlined for processing stock market reports into a predicate-argument representation of their content, for that part of the report which refers to the stock exchange activity. The procedure may have applications beyond information retrieval, in particular to the synthesis of informative stock market reports in one or more languages.

Embedded sublanguages and natural language processing

Most recent systems for the large-scale intelligent processing of natural language texts are designed to accept only a restricted variety of language. In certain cases this restricted subseto of the language constitutes a sublanguage , for which it may be possible to write a relatively precise and compact sublanguage grammar . Several research groups are currently exploiting the restrictions in scientific and technical sublanguage grammars for tasks such as information retrieval and automatic translation.