The important process of time series analysis for public health data is to determine target data as a semantic discrete value, according to a context from continuous phenomenon around our circumstance. Typically, each field of experts has their own fields’ specific and practical knowledge to specify an appropriate target part of data which contains the key features of their intended context in each analysis. Those are often implicit, thus not defined as systematically and quantitatively. In this paper, we present a context-based time series analysis and prediction method for public health data. The most essential point of our approach is to express a basis of time series context as the combination of the following 5 elements (1: granularity setting on time axis, 2: feature extraction method, 3: time-window setting, 4: differential computing function, and 5: pivot setting) to determine target data as semantic discrete values, according to the time series context of analysis for public health data. One of the main features of our method is to create different results by switching time series contexts. The method realizes 1) introducing a new normalization (context expression) method to fix a target reference data for time series analysis and prediction according to a context, and 2) presenting a process to generate semantic discrete values reflecting the 5 elements. And the significant features of the proposing method are 1) our context definition realizes the closed world of the semantic differential computing on time axis from the viewpoint of database system, and 2) the 5 elements enable to explicit and quantify experts’ semantic viewpoint of specifying a certain reference data according to a context for each analysis and prediction. As our experiment, we have realized analysis and prediction by applying actual public health data. The results of the experiments show the prediction feasibility of our method in the field of public health data, effectiveness to generate results for discussion regarding switching context, and applicability to express time series context of an expert knowledge for analysis and prediction as combination of the 5 elements to make the knowledge explicit and quantitative expression.

It is significant to detect, estimate and predict “Human-health situations” and “a spread of transmitted disease” with past and current information of health-related phenomena. Temporal-transition and differential computing realizes semantic interpretations for situation changes in two phenomena with “temporal-length” in “specific situation”. The “temporal-length” in “specific situation” is used to compare two phenomena in multiple contexts in semantics. We present a new Temporal-transition Differential Computing Model for detecting, estimating and predicting “Human-health situations” and “a spread of transmitted disease.” This model defines “temporal-transition data structure” for expressing past and current information of health-related phenomena with temporal-axis, and two processes for Human-Health Semantic Space Creation and Semantic Computing with dimensional control mechanism.

There has been significant progress in the field of sentiment analysis. However, aspect-based sentiment analysis (ABSA) has not been explored in the Japanese language even though it has a huge scope in many natural language processing applications such as 1) tracking sentiment towards products, movies, politicians etc; 2) improving customer relation models. The main reason behind this is that there is no standard Japanese dataset available for ABSA task. In this paper, we present the first standard Japanese dataset for the hotel reviews domain. The proposed dataset contains 53,192 review sentences with seven aspect categories and two polarity labels. We perform experiments on this dataset using popular ABSA approaches and report error analysis. Our experiments show that contextual models such as BERT works very well for the ABSA task in the Japanese language and also show the need to focus on other NLP tasks for better performance through our error analysis.

Semantic space creation and computing are essentially significant to realize semantic interpretations of situations and symptoms in human-health. We have presented a semantic space creation and computing method for domain-specific research areas. This method realizes semantic space creation with domain-oriented knowledge and databases. This paper presents a semantic space creation and computing method for “Human-Health Database” with the implementation process for “Human-Health-Analytical Semantic Computing”. This paper also presents a new knowledge base creation method for personal health data for preventive care and potential risk inspection with global and geographical mapping and visualization in 5-Dimensional World Map System. This method focuses on the analysis of personal health and potential-risk inspection and realizes a set of semantic computing functions for semantic interpretations of situations and symptoms in human-health. This method is applied to “Human-Health-Analytical Semantic Computing” to realize world-wide evaluation for (1) multi-parameterized personal health data, such as various biomarkers, clinical physical parameters, lifestyle parameters, other clinical/physiological or human health factors, etc., for health monitoring, and (2) time-series multi-parameterized health data in the national/regional level for global analysis of potential cause of disease. This Human-Health-Analytical Semantic Computing method realizes a new multidimensional data analysis and knowledge sharing for a global-level health monitoring and disease analysis. The computational results are able to be visualized in the time-series difference of the values in each place, the difference between the values of multiple places in a focused area, and the time-series differences between the values of multiple places to detect and predict a potential-risk of diseases.

This paper presents a new knowledge base creation method for personal/collective health data with knowledge of preemptive care and potential risk inspection with a global and geographical mapping and visualization functions of 5D World Map System. The final goal of this research project is a realization of a system to analyze the personal health/bio data and potential-risk inspection data and provide a set of appropriate coping strategies and alert with semantic computing technologies. The main feature of 5D World Map System is to provide a platform of collaborative work for users to perform a global analysis for sensing data in a physical space along with the related multimedia data in a cyber space, on a single view of time-series maps based on the spatiotemporal and semantic correlation calculations. In this application, the concrete target data for world-wide evaluation is (1) multi-parameter personal health/bio data such as blood pressure, blood glucose, BMI, uric acid level etc. and daily habit data such as food, smoking, drinking etc., for a health monitoring and (2) time-series multi-parameter collective health/bio data in the national/regional level for global analysis of potential cause of disease. This application realizes a new multidimensional data analysis and knowledge sharing for both a personal and global level health monitoring and disease analysis. The results are able to be analyzed by the time-series difference of the value of each spot, the differences between the values of multiple places in a focused area, and the time-series differences between the values of multiple locations to detect and predict a potential-risk of diseases.

Traditional automatic question generation often requires hand-crafted templates or sophisticated NLP pipelines. Such approaches, however, require extensive labor and expertise to morphologically analyze the sentences and create the NLP framework. Our works aim to simplify these labors. We conduct a contrastive experiment between two types of sequence learning: statistical-based machine translation and attention-based sequence neural network. These models can be trained end-to-end, and it can capture the pattern between the input sequence and output sequence, thus diminishing the need to prepare a sophisticated NLP pipeline. Automatic evaluation results show that our system outperforms the state-of-the-art rule-based system, and also excels in terms of content quality and fluency according to a subjective human test.

In order to organize the large number of products listed in e-commerce sites, each product is usually assigned to one of the multi-level categories in the taxonomy tree. It is a time-consuming and difficult task for merchants to select proper categories within thousands of options for the products they sell. In this work, we propose an automatic classification tool to predict the matching category for a given product title and description. We used a combination of two different neural models, i.e., deep belief nets and deep autoencoders, for both titles and descriptions. We implemented a selective reconstruction approach for the input layer during the training of the deep neural networks, in order to scale-out for large-sized sparse feature vectors. GPUs are utilized in order to train neural networks in a reasonable time. We have trained our models for around 150 million products with a taxonomy tree with at most 5 levels that contains 28,338 leaf categories. Tests with millions of products show that our first predictions matches 81% of merchants’ assignments, when “others” categories are excluded.

This paper proposes a new UGC-oriented language technology application, which we call experience mining. Experience mining aims at automatically collecting instances of personal experiences as well as opinions from vast amounts of user generated cont

We demonstrate a prototype of a call-to-Web session linkage system for sharing a Web application between caller and receiver. The system makes a session linkage between Web browsers based on call session information of SIP, which is an application-layer control protocol mainly used for IP telephony systems. After that, the system offers Web servers information for the linkage so that Web servers can provide a shared Web service easily. This system can even be uses by a browser of home appliances, which are difficult to upgrade because those linkage protocols consist of basic browser functions such as HTTP redirect and JavaScript.

This paper proposes a new UGC-oriented language technology application, which we call experience mining. Experience mining aims at automatically collecting instances of personal experiences as well as opinions from an explosive number of user generated contents (UGCs) such as Weblog and forum posts and storing them in an experience database with semantically rich indices. After arguing the technical issues of this new task, we focus on the central problem, factuality analysis, among others and propose a machine learning-based solution as well as the task definition itself. Our empirical evaluation indicates that our factuality analysis task is sufficiently well-defined to achieve a high inter-annotator agreement and our factorial CRF-based model considerably outperforms the baseline. We also present an application system, which currently stores over 50M experience instances extracted from 150M Japanese blog posts with semantic indices and is scheduled to start serving as an experience search engine for unrestricted users in October.

Our context-aware information delivery system enables information to be provided directly to the standby screen of a user’s mobile phone. The information appears on the standby screen only while the user context matches the information context due to a function that continuously monitors the behavioral response to user context, time, location, and reference history. We conducted a four-month trial of local information provision with over 800 mobile phone users participating. Approximately 30% of them.. "positively accepted" this information provision and most users actually utilized the received information. This shows that our approach is effective for mobile ad delivery.

In this paper, we describe a motion pattern acquisition system from time-series posture data for a robot. In order to extract reusable motion patterns without the target pattern information a priori, proposed system extracts the patterns observed frequently. The extracted motion patterns often include transition motion at first and last. In order to reduce the transition part, we propose common block extraction method. The similar common block is classified to the same category. Then, the proposed system obtains the average of each cluster as representative motion patterns. Experimental results show remarkable motion patterns extracted by proposed system from time-series posture data.