In this study, we suggest an emotional model (affect and feeling) for a communication agent in order to realize natural communication between a human and the agent. The suggestion model has six emotions: “anger,” “disgust,” “fear,” “sad,” “happiness,” and “surprise.” The model is constructed using fuzzy cognitive maps. The system can express a plurality of emotions simultaneously by mutually connecting the six feelings.
In this paper, we reconsider the behavior policy and the value estimation from the point of view of Bayesian approach in the reinforcement learning, to devise a new algorithm based on Prospect Theory. We realize that good behavior is selected by probability distribution criteria based on Bayesian estimation, and thereby it can achieve superior learning in terms of search efficiency than the conventional method. Estimated value distribution functions are represented by a beta distribution, and behavior selection policy is carried out by evaluating their mean and variance. Two parameters of this beta distribution consist of reward and weighted parameters of the now and the next state for each positive and negative one, then they are updated like Q-learning. Reinforcement learning becomes possible by being updated on the basis of the prospect theory in order to correspond to the state transition. Each initial probability distribution is a uniform distribution. It is revealed that an advantage of the proposed method is the breadth of its search in a discrete space path problem. It is also showed that applicability to more complicated problem by continuous space path search problem.
Faculty development (FD) program has been widely conducted in response to social concern in quality of higher education. We believe that analyzing of audiences’ state with engineering approaches will provide an applicative way to make a feedback to teachers. Video analysis takes advantages in applicability to practical due to their low cost and small requirements. Conventional methodologies, however, did not identify and/or implement essential image features to analyze audiences’ state under various conditions and therefore the availability of them was limited. In this study, we apply the pattern recognition framework for this task and propose an estimation system of state of audiences with convolutional neural networks (CNN), which can obtain effective image features through their learning process. Then, we focus on major issues pertaining to the exploration of CNN and assess the obtained image features. Our proposed system achieved audience detection performance of precision = 84.8% and recall = 61.8% and state estimation accuracy = 72.8% under various conditions. In addition, CNN showed good performance for this task under the situation that only limited numbers of training data were given. We confirmed from these results that CNN obtained essential image features for estimating state of audiences appropriately.
While statistical machine translation methods have been developed by using parallel corpus, a technical issue of collecting large amounts of good quality parallel sentence pairs has been raised.With recursive learning, which yields quantification of differences between sentences of one language and sentences of the other language by a statistical machine translation using the parallel corpus, a novel method of parallel corpus revision (clean-up) is proposed in this paper.By applying edit numbers to the sentence difference quantification, we show experimental results of the clean-up using Japanese-English patent parallel corpus.