In this paper, we propose a melody generation method using Variable-order Pitman-Yor Language Model (VPYLM) to compose more natural phrase than using an ordinal n-gram model. This method is easier to be extended by adding various kinds of restrictions than previous methods based on deterministic models. This language model reduces a sparseness problem. It learns longer phrases than the ordinal model. Furthermore, this model needs few memory because of the efficient learning process. A melody is generated by using Gibbs Sampler in the actual time. An experiment shows that our method can generate the melody as if to tie various length of a phrase in the learning data. We also show that our model can generate more natural melody than the ordinal model by an evaluation experiment.
The necessity of automatic text summarization has been recently increasing for helping people choose their necessary information. While various methods have been proposed, the usefulness of graph-based text summarization methods such as LexRank is recognized. LexRank computes the importance of sentences based on the idea of eigenvector centrality in a graph representation of sentences. This method uses the surface information to measure similarity among sentences such as cosine similarity, but does not use latent topic similarity. In this study, we propose a multi-document summarization method using a graph of sentences based on latent topics and show our method can summarize multiple documents with higher accuracy than LexRank through an experiment using DUC2004 task.
This paper proposes a method for investigating Kansei impressions of products using a genetic algorithm (GA). We analyze the relationship between consumers' overall impression and their response to individual components. In our study, fuzzy rules are obtained on the basis of the results of a questionnaire to determine this relationship. The rules are some general rules describing the reactions that determine the overall impression of a product. This study aims to discover useful knowledge about the product design from the rules obtained in the previous study. In our previous study, we investigated the factors that determine whether users want products and obtained some reliable rules by using a fuzzy C4.5 decision tree built on the basis of the questionnaire results. However, there are some constraints in the method in our previous study. All rules include the attribute assigned as the top node of the decision tree. Therefore, the rules' performance has limitations. Therefore, in this study, we use a GA in parallel and obtain rules. We verify the effectiveness of the proposed technique by an experiment. Moreover, we compare the rules using the proposed technique with those using the fuzzy C4.5 decision tree built in our previous study. The experimental results show that the performance of the rules using the proposed technique is improved.
The Project Management Office (PMO) is responsible for the oversight of all projects. It is an important issue for monitoring project accomplishments against established criteria and advising the governance council of status and issues. PMO would affect the planned benefits in any project. Therefore, the purpose of this research is to establish an evaluation model for PMO implementation and shows how to use qualitative and quantitative measurements of PMO indexes/criteria to evaluate and create a best implementation, as well as how to help project-manager use these indexes to build a best doing systems in criteria priorities for achieving project goals. So, this research proposes a hybrid MCDM model, combined DEMATEL technique with ANP and VIKOR methods for probing the PMO problems and how to evaluate and create a best implementation for achieving the aspired levels. Then an empirical case for PMO of information technology (IT) as an example is illustrated to show the usefulness of the proposed model in practices. Different from the past, in this paper we focused on one IT PMO from a company; according to their information to calculate the priorities among each dimensions and criteria, and then performance are calculated. In addition, 35 experts questionnaire are collected in 2012. The results show that there were interactive relations between all the criteria, where the dimension of processes is the most influential, on the contrary the tools is the least dimension; in the nine evaluation criteria, the most important criterion is “experts and technical capacity” (C8) in the stand point of experts and scholars, followed by“Develop and coach upper management sponsors” (C1) and “Train project managers” (C3).