Ryosuke Saga

Trends Recognition in Journal Papers by Text Mining

To recognize the trends in journal papers, this paper discusses a text mining method and its application. The method is based on combination of the conventional TF-IDF algorithm for document indexing and KIM analysis in marketing research. While TF (term frequency) can be clue for strength of topics and IDF (inverted document frequency) can be clue for bias of topics, recency in RFM analysis can be clue of vicissitude of topics. Applying the proposed method to trend analysis for the quality control journals in the Japanese society, this paper describes how the cross-tabulation of TF, DF and LA (last appearance) recognizes the research trends.

Hotel recommender system based on user's preference transition

This paper proposes a hotel recommender system based on sales records. Basic premise under the research is that the sales records include the users preference relations among hotels. The proposed system recommends hotels based on preference transition network when a user selects a hotel. This paper describes four steps procedure for building the preference transition network proposes in detail. The proposed recommender system is available for repeatable purchase without explicit product evaluation. The features of the prototype system are also illustrated.

Visualized Technique for Trend Analysis of News Articles

In order to visualize keyword trends in texts of news articles, this paper proposes a method named FACT-Graph by extending co-occurrence graph. The method uses four classes of keywords, considers three patterns of class transitions, and expresses three types of co-occurrence relationships between two analysis periods. Classes of keywords are characterized by the shapes of their nodes, the transition patterns of keyword classes are shown by the colors of the nodes, and the co-occurrences relationships between two keywords are represented by the types of edges their nodes have. FACT-Graph is applied to a sample of 220,000 newspaper articles and is found to be effective in visualizing keyword trends embedded in volumes of text data.

FACT-Graph: Trend Visualization by Frequency and Co-occurrence

In order to visualize keyword trends embedded in documents, this paper proposes FACT-Graph (Frequency and Co-occurrence-based Trend Graph). First, we introduce four classes of keywords by TF (Term Frequency) and DF (Document Frequency). Then while some keywords stay in the same class between two periods, others stay in the difference classes. Paying attention to such class transition between periods, we make it a clue of trend analysis. Next, we identify relationship between keywords by their co-occurrence strength and their transition between two periods. Then, we propose FACT-Graph by combining class transition information and co-occurrence transition information. Finally, an application to newspaper article is also discussed.

Development and case study of trend analysis software based on FACT-Graph

This article proposes text mining software to analyze FACT-Graph, and describes a case study using the software. FACT-Graph is a trend graph which visualizes what kinds of topic exist and shows the changes in trends in time-series text data. However, FACT-Graph itself does not have enough environments to analyze trends although it provides clues for a trend. In order to resolve this problem, we developed the software called Loopo. This software provides the functions of adding the considerations of the analyst as the keywords, and operating FACT-Graph itself such as moving, adding, and clearing nodes. The system also allows analysts to refer to an information source, keyword information, and network information in order to analyze and consider FACT-Graph. In a case study about criminal trends using the titles of newspaper articles between 1987 and 2007, we confirmed the usability of this software.

Multi-type Edge Bundling in Force-Directed Layout and Evaluation

Abstract Numerous information visualization techniques are available for utilizing and analyzing large data. Among these techniques, network visualization, which employs node-link diagrams, can determine the relationship among multi-dimensional data. However, when data become extremely large, visualization becomes obscure because of visual clutter. To address this problem, many edge bundling techniques have been proposed. However, although graphs present several edge types, previous techniques do not reflect these edges. In this paper, we propose a new edge bundling method for multi-type co-occurring graphs. In this method, electro-static forces work between each pair of edges; however, if the edges are of different types, then repulsion works between pairs. By bundling edges of the same type, a graph can more clearly show relationships among data. Qualitative evaluation through questionnaires lead to useful knowledge, i.e., the proposed method improves bundling performance more extensively than other related work.

Improved Keyword Extraction by Separation into Multiple Document Sets According to Time Series

This study proposes a method of extracting keywords including those that appear locally. Useful keyword extraction methods are available for text mining, such as TF-IDF and support vector machine. However, when keywords are extracted on the basis of time series, the local keywords are not often extracted. We propose a method of extracting the local keywords by separating a document set, which we call the document separation approach. The approach splits a document set into multiple sets according to time series, extracts the keywords for each document set, and integrates them. Using 1812 newspaper articles, we experimentally demonstrate that we can extract the local feature keywords using the document separation approach.

FML-based prediction agent and its application to game of Go

In this paper, we present a robotic prediction agent including a darkforest Go engine, a fuzzy markup language (FML) assessment engine, an FML-based decision support engine, and a robot engine for game of Go application. The knowledge base and rule base of FML assessment engine are constructed by referring the information from the darkforest Go engine located in NUTN and OPU, for example, the number of MCTS simulations and winning rate prediction. The proposed robotic prediction agent first retrieves the database of Go competition website, and then the FML assessment engine infers the winning possibility based on the information generated by darkforest Go engine. The FML-based decision support engine computes the winning possibility based on the partial game situation inferred by FML assessment engine. Finally, the robot engine combines with the human-friendly robot partner PALRO, produced by Fujisoft incorporated, to report the game situation to human Go players. Experimental results show that the FML-based prediction agent can work effectively.

Edge Bundling in Multi-attributed Graphs

Numerous information visualization techniques are available for utilizing and analyzing big data. Among which, network visualization that employs node-link diagrams can determine the relationship among multi-dimensional data. However, when data become extremely large, visualization becomes obscure because of visual clutter. To address this problem, many edge bundling techniques have been proposed. However, although graphs have several attributions, previous techniques do not reflect these attributions. In this paper, we propose a new edge bundling method for attributed co-occurrence graphs. Electrostatic forces work between each pair of edges; however, if the edges are under different attributions, then repulsion works between pairs. By bundling edges under the same attribution, a graph can more clearly show the relationships among data.

Measurement Evaluation of Keyword Extraction Based on Topic Coverage

This paper proposes a method to measure the performance of keyword extraction based on topic coverage. The answer set of a keyword is required to evaluate keyword extraction by methods such as TF-IDF. However, creating an answer set for a large document is expensive. Thus, this paper proposes a new measurement called topic coverage on the basis of the assumption that the keywords extracted by a superior method can express the topic information efficiently. The experiment using the proceedings of a conference shows the feasibility of our proposed method.