No abstract

This study characterized focused tones in Mandarin Chinese through a production experiment using phone number strings. The results revealed that, although phonation cues had little effect on any focused tone, prosodic cues exhibited various patterns of distribution. Duration played an important role for each focused tone, but intensity had a relatively less salient role. Among pitch-related parameters, the raising of pitch register was an important cue when a level tone (tone 1) was focused. By contrast, due to the interaction between tone and intonation, absolute slope and pitch range had less effect on tone 1 focus. These cues, however, were prominent when contour tones (tones 2 and 4) were in focus. Unlike other focused tones that raised pitch, tone 3 focus exhibited the opposite pattern, lowering its pitch target. In the aggregation of all focused tones, it was found that only primarily pitch-related parameters were selected as the main variables discriminating one from another. The results of this study, therefore, suggest that the prosodic marking of focus is not uniform, even within a single language, but clearly differs by tone type. Accordingly, prosodic marking of focus should be considered multimodal in a tonal language.

It is known that VP-ellipsis and VP-anaphora are typologically different phenomena. English has VP-ellipses whereas Korean has VP-anaphora. The goals of this paper are (i) to develop a unified algorithm which can analyze these two different phenomena and (ii) to explain them using the developed resolution algorithm. In order to analyze these phenomena, this paper incorporates Jager's anaphora resolution mechanism (2010) into the typed feature structure formalism of Head-driven Phrase Structure Grammar (HPSG). In this paper, VP-ellipsis and VP-anaphora are analyzed as follows. First, English do and Korean kuleha-ta are introduced with the Geach value, and this value is changed with a slash-elimination rule. Then, one constituent combines with another by ordinary syntactic rules, while the information on the target predicate is percolated up. When a potential source appears, a slash-introduction rule is applied. Then, the source predicate activates the VP-resolution rule, and the target predicate is connected with the source in the semantic representations.