The Beckman Institute Graduate Student Seminar Series presents the work of outstanding graduate students working in Beckman research groups. The seminar begins at Noon in Beckman Institute Room 1005 and is open to the public. Lunch will be served.
Inverse Design for Phase Mask Lithography
Phase mask lithography (PML) has emerged as a promising route to the fabrication of three-dimensionally (3D) periodic micro- and nano-structures. However, efforts to employ PML in various fields have relied almost exclusively on trial and error design. To realize the full potential of this technique, a design strategy based on the use of genetic algorithms (GAs) has been formulated. This approach has been successfully implemented to design various phase masks that produce a given 3D structure. In this talk, the GA-based design method will be discussed and recent efforts to experimentally demonstrate these designs will be presented.
Automatic Assessment Method of Spontaneous Speech Fluency
This study aims at developing a method for the automatic assessment of spontaneous speech fluency. In previous studies, the automatic fluency assessment methods have focussed on the temporal aspect of speech especially disfluency. Cucchiarini et al. (2002) showed that in read speech, L2 speakers' proficiency level can be predicted by measures related to disfluency. However, they found that the correlation between the proficiency score and the disfluency measures was considerably weaker in spontaneous speech. Therefore, measures which can capture the characteristics of spontaneous speech are required. This study focuses on measures related to English rhythm. English is a stress-timed language, and the reduction of unstressed syllables is an important characteristic of English. In addition, English speakers reduce words more when the words are more predictable (Jurafsky et al., 2001). Measures related to these English rhythmic patterns are proposed and the proficiency level of L2 learners is predicted based on the disfluency and rhythmic measures. The method is trained and tested using a corpus of English spontaneous speech, collected from L2 learners of varying degrees of proficiency and a variety of L1's.