Thesis (M.S.) PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you. / Children with ASD suffer from language deficits that are in part due to impaired social abilities. In the current study, we test a non-social and noninteractive method for teaching word meanings. If successful, this technique will expand the ways we teach early vocabulary that would shape intervention and improve the quality of life and the future prospects of children with ASD. First, the participants (ages 26.09 - 47.37) listen to a series of transitive or intransitive sentences while watching an unrelated silent animation. These sentence types carry different meanings; transitive verbs describe causative actions, and intransitive verbs describe synchronous actions. Then they are asked to map the verb to meaning by finding the relevant action out of two candidate scenes. We analyze the children’s direction of gaze. The results revealed that children with ASD in this sample who heard transitive sentences in maximally non-social and non-interactive contexts did not show a significant preference for the causative scene in either test condition. Additional data will reveal whether children with ASD can learn from non-social presentation of novel words if given multiple exposures to the test scenes. This work will provide insight into the mechanisms underlying word learning in children with ASD and, if successful, provide the foundation for research into a new kind of non-social intervention that capitalizes on these children’s strengths. / 2031-01-01

Thesis (Ed. D.)--West Virginia University, 2007. / Title from document title page. Document formatted into pages; contains viii, 175 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 155-161).

This thesis investigated the use of novel combinations of multimodal metaphors in the presentation of learning information to evaluate the effect of these combinations on the usability of e-learning interfaces and on the users' learning performance. The empirical research described in this thesis comprised three experimental phases. In the first phase, an initial experiment was carried out with 30 users to explore and compare the usability and learning performance of facially animated expressive avatars with earcons and speech, and text with graphics metaphors. The second experimental phase involved an experiment conducted with 48 users to investigate their perception of avatar's facial expressions and body gestures when presented in both the absence and presence of interactive e-learning context. In addition, the experiment aimed at evaluating the role that an avatar could play as virtual lecturer in e-learning interfaces by comparing the usability and learning performance of three different modes of interaction: speaking facially expressive virtual lecturer, speaking facially expressive full-body animated virtual lecturer, and two speaking facially expressive virtual lecturers. In the third phase, a total of 24 users experimentally examined a novel approach for the use of earcons and auditory icons in e-learning interfaces to support an animated facially expressive avatar with body gestures during the presentation of the learning material. The obtained results demonstrated the usefulness of the tested metaphors to enhance e-learning usability and to enable users to attain better learning performance. These results provided a set of empirically derived innovative guidelines for the design and use of these metaphors to generate more usable e-learning interfaces. For example, when designing avatars as animated virtual lecturers in e-learning interfaces, specific facial expression and body gestures should be incorporated due to its positive influence in enhancing learners' attitude towards the learning process.

While recent neural sequence-to-sequence models have greatly improved the quality of speech synthesis, there has not been a system capable of fast training, fast inference and high-quality audio synthesis at the same time. In this the- sis, we present a neural speech synthesis system capable of high-quality faster- than-real-time spectrogram synthesis, with low requirements on computational resources and fast training time. Our system consists of a teacher and a student network. The teacher model is used to extract alignment between the text to synthesize and the corresponding spectrogram. The student uses the alignments from the teacher model to synthesize mel-scale spectrograms from a phonemic representation of the input text efficiently. Both systems utilize simple convo- lutional layers. We train both systems on the english LJSpeech dataset. The quality of samples synthesized by our model was rated significantly higher than baseline models. Our model can be efficiently trained on a single GPU and can run in real time even on a CPU. 1

Cochlea-Implantate sind Neuroprothesen, die es Gehörlosen ermöglichen, Zugang zu auditiver Information wieder zu erlangen. Allerdings ist das resultierende Signal stark verzerrt bzw. degradiert und eine erfolgreiche Adaptation oft unvollständig. Das Verständnis zugrundeliegender perzeptueller Lernprozesse ist somit von enormer klinischer Bedeutung. Perzeptuelles Lernen degradierter Sprache lässt sich mittels Noise-Vokodierung bei Hörgesunden simulieren. In Bildgebungsstudien konnte anhand funktioneller Magnetresonanztomographie gezeigt werden, dass perzeptuelles Lernen degradierter Sprache mit einer Aktivitätssteigerung im linken inferior frontal gyrus (IFG) und linken inferior parietal cortex (IPC) assoziiert war. Die vorliegende Studie untersucht den Einfluss von fazilitierender, nicht-invasiver Hirnstromstimulation (anodale transkranielle Gleichstromstimulation, tDCS) über dem linken IFG und linken IPC auf das perzeptuelle Lernen {\\itshape{Noise}}-vokodierter Sprache. Die Probanden trainierten die Diskrimination von Minimal- (\"Tisch\"- \"Fisch\") und identen (\"Tisch\"- \"Tisch\") Wortpaaren, während der erste Stimulus akustisch degradiert, der zweite in geschriebener Form präsentiert wurden. Vor und nach dem Training wurden die trainierten Stimuli und eine gleiche Anzahl untrainierter Stimuli präsentiert. Perzeptuelles Lernen wurde in unserer Studie als eine Verbesserung der Diskriminationsleistung untrainierter Wortpaare operationalisiert. Zudem wurde vor und nach dem Training ein Elektroenzephalogramm abgeleitet. Auf elektrophysiologischer Ebene wurde der Einfluss des Lernvorgangs und der tDCS auf die N400 untersucht, welche mit der Verarbeitung lexiko-semantischer Informationen assoziiert ist. Unsere Ergebnisse zeigen, dass eine anodale tDCS über dem linken IFG perzeptuelles Lernen stark degradierter Sprache fazilitierte, während bei einer Placebo- bzw. einer Stimulation über dem linken IPC kein perzeptuelles Lernen stattfand.