The aims of this research were to develop guidelines for designing interaction tasks for learners of Chinese as a foreign language (CFL) and to investigate the attitudes of CFL learners toward a full CFL class in Second Life (SL). Three research questions were addressed in this research: (1) what are the attitudes of CFL learners toward the language learning tasks in SL? (2) what kinds of social interactions emerge from learning activities in a CFL class in SL? (3) how do those activities benefit CFL learners in the learning of Chinese in SL? Two studies were conducted to tackle these questions. The cognition, usage, and expansion (CUE) model was proposed based on the findings obtained from study 1 and then implemented and evaluated in study 2. The findings of study 2 indicated that the activities run in the CUE model were effective at motivating CFL beginners and improving their oral communication and social interactions. Based on the video data analysis, three criteria were proposed for designing learning activities. Suggestions are also made for future research on CFL teaching/learning in SL.

No abstract

Incomplete multi-view clustering (IMVC) methods have attracted extensive attention in the field of data mining and pattern recognition due to their superior performance in processing incomplete multi-view data (IMVD). However, the existing IMVC methods can only obtain one consensus matrix by using only one fusion strategy, i.e., fusing the raw IMVD information, and then use k-means or spectral clustering to obtain final results. Although these methods achieve good performance, these algorithms have the following limitations. First, only one consensus matrix may not fully express the information of the IMVD. Second, the obtained consensus matrix may contain noise. Therefore, we propose an innovative method, i.e., incomplete multi-view clustering with a dual fusion strategy (IMVC/DFS) in this paper. In the first fuse process, IMVC/DFS considers how to fuse the raw IMVD information. Specifically, IMVC/DFS first converts the feature matrices into similarity graph matrices and uses an adaptive graph completion strategy to complete missing data. Different from the existing algorithms, IMVC/DFS does not obtain only one consensus matrix, but selects multiple parameters and uses tensor low-rank constraint and consensus representation term to obtain multiple consensus matrices. In the second fuse process, IMVC/DFS considers how to fuse the information contained in the different consensus matrices. Then, motivated by the self-expressiveness property of data, IMVC/DFS further obtain the self-representation matrices of all the consensus matrices and remove noise simultaneously. Finally, we fuse the self-representation matrices for clustering, which is the second fusion. According to our knowledge, IMVC/DFS is the first dual fusion method to solve the IMVC problem. We exploit Augmented Lagrange Multiplier (ALM) method to deal with IMVC/DFS, and the experimental results are very promising and fully demonstrate the validity and superiority of IMVC/DFS.