Computer‐supported collaborative learning (CSCL) has shown considerable promise, but thus far the literature has tended to focus on individual technological tools, without due regard for how the choice of one such tool over another impacts CSCL, either in outline or in detail. The present study, therefore, directly compared the learning‐related uses of an online discussion forum against such use of a mobile instant‐messaging app by the same group of 78 upper‐division undergraduate pre‐service teachers in China. The participants were asked to use one of the two communication tools during the first of three learning activities, then to switch to the other during the second, and to choose their preferred tool for the third. Based on the results of content analysis, social‐network analysis and a survey of the students' attitudes, it was found that while both tools facilitated collaborative learning, they appeared to have different affordances. Specifically, using the online discussion forum resulted in more communication aimed at knowledge construction, while using the mobile instant‐messaging app resulted in more social interactions.

Several consumer speech devices feature voice interfaces that perform on-device keyword spotting to initiate user interactions. Accurate on-device keyword spotting within a tight CPU budget is crucial for such devices. Motivated by this, we investigated two ways to improve deep neural network (DNN) acoustic models for keyword spotting without increasing CPU usage. First, we used low-rank weight matrices throughout the DNN. This allowed us to increase representational power by increasing the number of hidden nodes per layer without changing the total number of multiplications. Second, we used knowledge distilled from an ensemble of much larger DNNs used only during training. We systematically evaluated these two approaches on a massive corpus of far-field utterances. Alone both techniques improve performance and together they combine to give significant reductions in false alarms and misses without increasing CPU or memory usage.

No abstract

For real-world speech recognition applications, noise robustness is still a challenge. In this work, we adopt the teacherstudent (T/S) learning technique using a parallel clean and noisy corpus for improving automatic speech recognition (ASR) performance under multimedia noise. On top of that, we apply a logits selection method which only preserves the k highest values to prevent wrong emphasis of knowledge from the teacher and to reduce bandwidth needed for transferring data. We incorporate up to 8000 hours of untranscribed data for training and present our results on sequence trained models apart from cross entropy trained ones. The best sequence trained student model yields relative word error rate (WER) reductions of approximately 10.1%, 28.7% and 19.6% on our clean, simulated noisy and real test sets respectively comparing to a sequence trained teacher.Index Termsautomatic speech recognition, noise robustness, teacher-student training, domain adaptation * Ladislav Mosner performed the work while he was a research intern at Amazon.

No abstract

No abstract

The use of new technology encouraged exploration of the effectiveness and difference of collaborative learning in blended learning environments. This study investigated the social interactive network of students, level of knowledge building and perception level on usefulness in online and mobile collaborative learning environments in higher education. WeChat, which is a mobile synchronous communication tool, and modular object-oriented dynamic learning environment (Moodle) were used as mobile and online collaborative learning settings. Seventy-eight college students majoring in information engineering participated in the experiment. The following findings were revealed by combining methods of social network analysis, content analysis and questionnaire survey: (1) the collaborative social networks generated in this study showed that students had tighter interaction relationships in Moodle than in WeChat; (2) deeper level of knowledge building in collaboration and interaction through Moodle than WeChat was observed; and (3) Moodle got higher perception level than WeChat because of its usefulness for collaboration.

No abstract