Pronouncing Text by Analogy

Pronouncing Text by Analogy Robert I. Damper and John EG. Eastmond Image, Speech and Intelligent Systems (ISIS) Research Group, Department of Electronics and Computer Science, University of Southampton, Southampton SO17 IBJ, UK. {ridlje}@ecs. soton, ac.uk Abstract Pronunciation-by-analogy (PbA) is an emer- ging technique for text-phoneme conversion based on a psychological model of read- ing aloud. This paper explores the impact of certain basic implementational choices on the performance of various PbA mod- els. These have been tested on their abil- ity to pronounce sets of short pseudowords previously used in similar studies, as well as lexical words temporarily removed from the dictionary. Best results of 85.7% and 67.9% words correct are obtained lor the pseudo- words and lexical words respectively, cast- ing doubt on certain previous-reported per- formance figures in the literature. 1 Introduction Pronunciation-by-analogy (PbA) is an influential psy- chological model of the process of reading aloud. In PbA, most words are pronounced by retrieving their phonemic form from the readers's lexicon, or diction- ary. The pronunciation for a 'novel' word not in the lexicon, however, is derived not by the application of abstract letter-to-sound rules hut is 'assembled' from the (known) pronunciations of words that it resembles. PbA has obvious application to text-to-speech conver- sion by machine. Although PbA programs have been presented in the literature, they are they are few in number. Ded- ina and Nusbaum (1991) describe PRONOUNCE: a rather simple system for English. Sullivan and Damper (1990; 1992; 1993) describe a considerably more complex and developed system, but which apparently yields a much poorer perfornmnce. As a psychological theory, PbA is under-specified: offering little meaningfifl guidance on the implement- ation choices which confront the programmer. Indeed, Sullivan and Damper (1993) show that such choices can have a profound impact on performance. In this paper, we seek to understand how Dedina and Nus- baum's largely unjustified implementational choices affected their results and, thereby, to resolve the con- flict between their performance claims and Sullivan and Damper's. 2 Psychological Background In the standard dual-route model of reading aloud (Coltheart, 1978), there is a lexical route for the pro- nunciation of known words and a parallel route util- ising abstract letter-to-sound rules for the pronunci- ation of unknown ('novel') words. Arguments for dual-route theory cite the ability to pronounce pseudo- words (non-words conforming to the spelling patterns of English), latency difference effects between regular and exception words, and apparent double dissociation between the two routes in dyslexia (see Humphreys and Evett, 1985). However, all these observations can arguably be explained by a single route. One pervasive idea is that pseudowords are pronounced by analogy with lexical words that they resemble (Baron, 1977; Brooks, 1977; Glushko, 1979; 1981; Brown and Be- sner, 1987). Glushko, for instance, showed that "ex- ception pseudowords" like tave take longer to read than "regular pseudowords" such as taze. Here, taze is considered as a "regular pseudoword" since all its orthographic 'neighbours' (raze, gaze, maze etc.) have the regular vowel pronunciation/el/. By contrast, tave is …

Content truncated to 500 words