Working with analogical semantics: Disambiguation techniques in DLT By Victor Sadler (review)

BOOK NOTICES 419 various factors influencing xenolect pattern variation, and (3) to discuss the causes and functions of xenolect speech based on the results and hypotheses of the present study. Chs. 3-5 (20-137) present the main body of R's research. He outlines his methods of data collect· ?n and processing and describes in detail the different data pools and the subjects involved in each. In his analysis he postulates four different patterns or levels of utterance ('Außerungsstufen' a-d) and ascribes various characteristics to each level, concentrating ultimately on the d-level, the one that exhibits the most change with respect to the standard language . This d-level serves as the basis of R's attempt to establish a structure for xenolects. Developing his own terminology throughout. R then undertakes the task of working out a full structural system for xenolects, moving from the single-utterance level through a sequence of utterances, and dealing finally with an entire narrative text. Along the way he treats such topics as personal, temporal, and spatial reference , the scope of quantification, gradation, modality, negation, and confirmation, and lexicosemantic simplification. Having posited this ideal structural model, R turns his attention in Chs. 6-7 (138-76) to variations that can occur in the ideal structural system, a topic that he considers untrodden ground. He develops his own method of subjective evaluation of the language proficiency level of nonnative speakers, then examines his data to determine the degree of influence of these nonnative proficiency levels on the xenolect patterns of native speakers. R posits xenolect text structure as a series of text-pragmatic stages ('textpragmatische Niveaus') for which the speaker chooses what he feels to be the appropriate utterance level (a-d) to maximize communicative effectiveness. A definition of each of these proposed stages closes Ch. 7. In the final two chapters, R presents a general summary of his findings and their relevance for the field and concludes with a concise restatement of the goals and methods of his research. He also touches briefly on the possible relevance of his findings for disciplines tangential to xenolect studies. Through constant references to past studies in the field (albeit to lament their methods and refute their results) as background to his own original argumentation, R has produced a useful survey of past contributions as well as an important new contribution to the field. Moreover, this study should prove valuable not only for xenolectology, but also for such related fields as second-language acquisition and aphasia. The methodical presentation of the arguments, data, and analyses and the rich use of data samples and clear, illustrative graphics throughout the text make it accessible even to a beginner in the field. [Mark R. Lauersdorf, University of Kansas.] Working with analogical semantics: Disambiguation techniques in DLT. By Victor Sadler. (Distributed language translation, 5.) Dordrecht & Providence: Foris, 1989. Pp. 256. Cloth Dfl 120.00, paper DfI 67.00. The meaning of a word is its use in the language . This is a recurrent idea which amounts to the assertion that words with similar meanings have similar collocations. Before the computer age, collocational analysis ofmeaning was plausible but impractical; Sadler's book attempts to demonstrate that it can now be made the keystone of a workable machine-translation system. Distributed Language Translation is 'distributed ' in the sense that it is designed to distribute automatic machine translations to the users of a computer network. Such a system imposes special constraints: in many areas (e.g. Europe) the users will not speak the same language , and it will not be possible to check the quality of translations after transmission to the user. Many of the characteristics of S's system follow from these requirements, especially the choice of Esperanto as an interlanguage. S argues that Esperanto's limited lexicon of roots and its highly productive, regular morphology permit an intermediate representation that minimizes ambiguity, makes semantic relations (relatively) explicit, and is therefore (in principle ) appropriate for automatic translation without human revision. More practically, the goal is to achieve an unambiguous translation from the source language into Esperanto (through a user-machine dialog system) which can then serve as the source for automatic...