Humans develop their concept of an object by classifying it into a category, and acquire language by interacting with others at the same time. Thus, the meaning of a word can be learnt by connecting the recognized word and concept. We consider such an ability to be important in allowing robots to flexibly develop their knowledge of language and concepts. Accordingly, we propose a method that enables robots to acquire such knowledge. The object concept is formed by classifying multimodal information acquired from objects, and the language model is acquired from human speech describing object features. We propose a stochastic model of language and concepts, and knowledge is learnt by estimating the model parameters. The important point is that language and concepts are interdependent. There is a high probability that the same words will be uttered to objects in the same category. Similarly, objects to which the same words are uttered are highly likely to have the same features. Using this relation, the accuracy of both speech recognition and object classification can be improved by the proposed method. However, it is difficult to directly estimate the parameters of the proposed model, because there are many parameters that are required. Therefore, we approximate the proposed model, and estimate its parameters using a nested Pitman--Yor language model and multimodal latent Dirichlet allocation to acquire the language and concept, respectively. The experimental results show that the accuracy of speech recognition and object classification is improved by the proposed method.
Predicting human activities is important for improving recommender systems or analyzing social relationships among users. Those human activities are usually represented as multi-object relationships (e.g. user's tagging activities for items or user's tweeting activities at some locations). Since multi-object relationships are naturally represented as a tensor, tensor factorization is becoming more important for predicting users' possible activities. However, its prediction accuracy is weak for ambiguous and/or sparsely observed objects. Our solution, Semantic data Representation for Tensor Factorization (SRTF), tackles these problems by incorporating semantics into tensor factorization based on the following ideas: (1) It first links objects to vocabularies/taxonomies and resolves the ambiguity caused by objects that can be used for multiple purposes. (2) It then links objects to composite classes that merge classes in different kinds of vocabularies/taxonomies (e.g. classes in vocabularies for movie genres and those for directors) to avoid low prediction accuracy caused by rough-grained semantics. (3) It finally lifts sparsely observed objects into their classes to solve the sparsity problem for rarely observed objects. To the best of our knowledge, this is the first study that leverages semantics to inject expert knowledge into tensor factorization. Experiments show that SRTF achieves up to 10% higher accuracy (lower RMSE value) than state-of-the-art methods.
We developed a learning environment to support participants' problem posing in a formal logic system, natural deduction, by combining problem-posing and problem-solving activities. In the problem posing-phase, the participants posed original problems and presented them on a shared problem database called ``Forum,'' which was accessible to other group members. During the problem-solving phase, the participants solved the problems presented on Forum. This first round of problem posing and solving was followed by a second round of problem posing. We performed two practices: one for undergraduates in a liberal arts college and the other for graduates in a graduate school of information science. The results showed that the participants successfully posed more advanced problems in the second round of problem posing as compared to the first. The empirical data gathered from the two practices indicated a significant relationship between problem-solving and problem-posing abilities.
When people understand an object, they construct a mental model of the object. A mental model is a structural, behavioral, or functional analog representation of a real-world or imaginary situation, event, or process. We conducted a class practice in which newcomers to cognitive science constructed a mental model by implementing and simulating a computational model of cognitive information processing, i.e., a cognitive model. We quantitatively evaluated the learning outcomes of the class. The participants were required to implement a complete cognitive model of subtraction processing. Furthermore, they were required to implement bug models, which are cognitive models with bug rules that cause several types of errors. Pre- and post-tests were performed before and after implementing and using these models, respectively. The results indicate that the class intervention led to the increase of the number of the participants who constructed the correct mental model and promoted more accurate mental simulations. However, the significant effects were confirmed only with participants who correctly completed the bug model, but the effects were limited with those who failed.
Data interpretation based on theory is one of most important skills in scientific discovery learning, but to achieve this process is difficult for learners. In this study, we propose that model construction and execution could support data interpretation based on theory. We used the web-based production system ``DoCoPro'' as an environment for model construction and execution, and we designed and evaluated class practice in cognitive science domain to confirm our ideas. Fifty-three undergraduate students attended the course in Practice 1 in 2012. During class, students constructed a computational model on the process of semantic memory and conducted simulations using their model from which we evaluated any changes in learner interpretation of experimental data from pretest to posttest. The results of comparing pretest with posttest showed that the number of theory-based interpretations increase from pretest to posttest. However, we could not confirm the relationship between students' interpretations and their mental models acquired through learning activities and whether the students could transfer their understanding of theory to other different experimental data. Therefore, we conducted Practice 2 in 2013, in which 39 undergraduate students attended the course. Instruction in Practice 2 was same as in Practice 1. We improved pretest and posttest to assess students' mental model of theory and whether they transfer their understanding to another experiment. Comparing the pretest and posttest results showed that students acquired more sophisticated mental models from pretest to posttest, and they could apply their understanding of theory to their interpretations of near transfer experimental data. The results also indicated that students who shifted their interpretations from non theory-based to theory-based acquired more superior mental models on theory. Finally, we discuss applicability of our findings to scientific education.
This paper addresses the issue of how to develop skill in operating cognitive tool for learning. Cognitive tool is an interactive application for encouraging learners to visualize/externalize the process or results of learning and to scaffold the learning process as modeled. Such scaffolding enables them to gain cognitive experience of the learning process. Although the learning skill development generally requires the learners to accumulate the cognitive experiences in operating the tool, it would not always induce them to reflect on how they operate it and on how they learn via the tool. Such reflection contributes to learning how to operate the tool, which also involves learning how to learn. In order to provide learners with an opportunity for reflection, this paper proposes fadable scaffolding with cognitive tool, in which functions available on the tool can be faded according to learning skill in a learner-adaptable way. Such fading enables learners not only to accomplish the learning process without the tool but also to gain a proper and deeper understanding of the functions to become more skillful in operating the tool. This paper demonstrates fadable scaffolding with Interactive History (IH for short) for developing skill in operating it, which is a cognitive tool for scaffolding navigation and knowledge construction process in hyperspace provided with unstructured and hypertext-based resources. We have conducted a case study with the fadable scaffolding with IH. The results suggest that the fadable scaffolding allows learners to fade the IH functions in a reasonable way and to become more skillful in operating IH, which would contribute to become more skillful in navigation and knowledge construction process.
In order to facilitate learners' knowledge refinement process, it is effective to let them externalize their knowledge. However, in a domain of the instructional design in which existence of knowledge and its necessity are not sufficiently articulated or recognized, it is not easy for teachers who are also learners of how to externalize their knowledge. In this study, we have built a system called ``FIMA-Light'' which uncovers knowledge that teachers must have applied in their lesson plans from global to local viewpoints instead of them. FIMA-Light makes use of the OMNIBUS ontology which describes various instructional knowledge for attaining educational goals extracted from instructional/ learning theories. And, FIMA-Light automatically generates what we call I_L event decomposition trees by interpreting a given lesson plan based on the OMNIBUS ontology. Then, FIMA-Light facilitates teachers' deep reflection and helps them to refine their lesson plans by providing them with decomposition trees. We report some results of an experiment carried out for evaluation of the quality and the effectiveness of I_L event decomposition trees.
For skill acquisition that needs periodic body movements as cascade juggling, the establishment of stable body movements seems crucial. We investigated them in each of the learning stages defined by the Beek and van Santvoord (1992) framework. In addition, we investigated participants' verbal reports about what was intentionally concerned for achieving optimum learning in practice. In the experiment, novices practiced three-ball cascade juggling over a period of one week. We focused on two types of stabilities: the stability of chest movement representing torso movement, and another stability of wrist movement representing arm swing. The result revealed that the skills for establishing stabilities of torso movement and arm swing were acquired sequentially. In this case, the stability of arm swing emerged between Stage 2 (by 50 successive catches) and Stage 3 (by over 100 successive catches), and another stability of torso movement emerged between Stage 3 and the expert stage in which jugglers had acquired complete skills for performing five-ball cascade juggling. The result also showed that in the establishment of stable arm swing, the development of the stability occurred only in passive catching behavior, but did not in active tossing behavior. Additionally, we found that the participants who did not develop beyond Stage 1 (by 10 successive catches) trained themselves while focusing on their specific physical movements.
This research examines effects of visualization of learners' selections of theme and partners in long-run learning situations. Recently, there increased the project-based learning that requires learners to choose tasks and partners according to their own interests. Although social network analysis is suitable for visualization and analysis of those choices, it is not yet clear if those analyses work for teachers in situ. This study developed a system for visualizing the bird's-eye view of social interaction and learning process. We analyzed "selections" in the free play of elementary school kids and those in the graduation thesis production of college students. Results showed that the both analyses contributed to enriching teachers' confirmation or revision of their rules of thumb on learning. Based on these case studies we propose a system prototype that condenses a large amount of process data of learners' selections and makes it easy for teachers to look back results with minimum load and to revise next lessons in a timely manner.