Semiautomatic Interactive Muitilingual Style Analysis (SIMSA)

Semiautomatic Interactive Multilingual Style Analysis (SIMSA) G/inter Winkelmann TA Triumph-Adler AG TA Research Ftirther Strasse 212 D-8500 Nuremberg 80 1 Abstract A style checker is a tool which supports authors during the process of writing: Certain style markers are analyzed, their values are compared with a given norm, deviations are detected, and recommendations are given to the author. The power of a style checker depends on available tools such as lexica, parser, etc. This paper describes a style checker which will be integrated m a workbench (Translator's Workbench) and which has access to lexica and parser. The style checker can be used for different languages and for different kinds of text. defined with the help of these values: A stylistic norm fl~r technical texts, for user manuals, for a certain author, etc. With respect to these premises we now define style checking as: matching the densities of the features in a given text against the densities of the corresponding features in the norm. "Correct style" is correct style concerning a certain (group based) norm. "Incorrect style" is the degree of deviation between the norm values of the style markers and the values in the actual text. 3 Recent Approaches 2 Introduction Within ESPRIT II - Project 2315, Translator's Workbench (TWB), a tool is under development which checks stylistic markers of texts. According to the goals of Translator's Workbench as an integrated multilingual toolkit, the concept of the style checker includes multilinguality, and it uses other tools of Translator's Workbench such as parser and lexicon, which provide SIMSA with more power than comparable approaches. Style as it is understood in SIMSA is not reduced to simply a personal impression. Its description is not restricted to informal classifications such as "good style", "style like the style of author x". Within a functional, group based definition, style is the selection of certain words, phrases, sentences or structures out of a set of grammatically correct words, phrases, sentences or structures. These linguistic features in which texts can differ stylistically are called style markers. Group based means that these style markers have equal characteristics within a certain group of texts. The stylistic characteristics of a group of texts, their values of style markers, can be set as a norm. Thus, different norms can be Similar approaches (but within different contexts) have been done during the last decade. Beside approaches translating style from one language to another via abstract universal categories like text complexity and readability (Dimarco/Hirst 1988), two approaches to style checking, Writer's Workbench and EPISTLE, should be mentioned. Writer's Workbench (Cherry 1983 et al.) has influenced the development of commercial software such as Rightwriter, PC-Style and Grammatik. It does style critiquing, but it cannot do critiquing that reqmres a parser output, such as noun phrase complexity. Embedded in EPISTLE (Heidorn et al. 1982) is a style checker which uses the parse tree of the grammar check. So, EPISTLE covers a wider range of style markers. Nevertheless, stylistic critiques have to be adapted to the …

Content truncated to 500 words