Constituent ordering in spoken Japanese
MetadataShow full item record
In speaking, the speaker has a general idea about what they are going to talk about. They know how to form a grammatical sentence, and they know what grammatical roles they are going to assign to who and what. In a language like Japanese where the assignment of grammatical roles does not involve constituent ordering, the speaker has to decide in which constituent ordering they deliver the sentence. For example, when Japanese speakers want to say "Fred saw John", they first assign case particles, possibly the nominative case particle ga for "Fred" and the accusative case particle o for "John". The speaker then has to decide in which order, Fred-ga John-o or John-o Fred-ga , they will deliver the sentence. I first investigate three factors that are pertinent to the choice of one constituent ordering over another: recency of previous mentions, cataphoric persistence and syntactic heaviness of the constituent. Recency is quantified by Referential Distance (RD), cataphoric persistence by Topic Persistence (TP), and syntactic heaviness by Syntactic Weight (SW). The relative measurement of RD, TP and SW between the two constituents in question indicates that the constituent with a lower RD, a higher TP or a higher SW tends to come before the constituent with a higher RD, a lower TP or a lower SW. Based on the patterns between factors and the earlier position, algorithms are formulated to examine if the algorithms can correctly predict the ordering choice between SOV and OSV. Applying the algorithms to the data shows not only that single predictors, RD, TP and SW, can make correct predictions, but also that the combination of three predictors can predict with a higher success rate than the single predictor or the combination of two predictors. In addition, other factors, the focus structure, the use of the particle wa and demonstratives, the light verb construction, and scene-setting and topical elements, are explored. The algorithms based on the factors are further applied to clauses containing the subject, the direct object and other phrases to examine if the factors are relevant to predicting what comes first or what precedes what.