人工知能学会論文誌
検索
OR
閲覧
検索
32 巻 , 2 号
選択された号の論文の6件中1~6を表示しています
    • |<
    • <
    • 1
    • >
    • >|
原著論文
  • 小林 潤平, 関口 隆, 新堀 英二, 川嶋 稔夫
    32 巻 (2017) 2 号 p. A-AI30_1-24
    公開日: 2017/03/01
    ジャーナル フリー

    We propose bunsetsu-based layouts to improve the efficiency of eye movements in reading Japanese text. When reading text, people tend to direct their gaze toward the center of a word. This is called the optimal viewing position. The optimal viewing position results in the shortest gaze durations and fewest re-fixations. In the case of Japanese text, the eyes tend to fixate on each characteristic Japanese linguistic unit (bunsetsu). An electronic Japanese text reader that facilitates accurate control of eye movements to bunsetsu segments could increase the reading rateand result in more efficient eye movements. In this study, we develop new techniques to decrease inefficient eye movements when reading Japanese text. In Experiment I, we investigate the effectiveness of the layout with bunsetsu-based line breaking. A bunsetsubased linefeed layout breaks a line between bunsetsu segments, i.e., splitting a bunsetsu segment is prohibited. The reading speed for the bunsetsu-based linefeed layout was faster compared to the conventional text layout with line lengths of 5–40 characters per line. The improvements in reading speed were likely due to the optimization of eye movements near the edge of a line. In the case of 5–11 characters per line, the improvements in reading speed were likely due to an increase in the number of lines that can be recognized by a single fixation. These results indicate that the bunsetsu-based linefeed layout is an effective technique to improve reading efficiency. In Experiment II, we develop a new micro-vibration text reader with bunsetsu-based segmentation. The new reader vibrates each bunsetsu segment in a different phase to enhance boundary information for eye guidance. The reading speed for the micro-vibration text was approximately 7%–12% faster compared to the stable text with line lengths of approximately 11–29 characters per line. The improvements in reading speed were likely due to a reduction in re-fixations within a bunsetsu segment and an increase in the number of lines that can be recognized by a single fixation without horizontal saccades. Moreover, 76% of the participants did not experience illegibility or incongruousness with the micro-vibration text reader. These results indicate that micro-vibration is an effective technique to improve the efficiency of reading text with line lengths of 11–29 characters per line without an increase in cognitive load or a decrease in comprehension. In Experiment III, we develop a new stepwise incremental indent layout with bunsetsu-based segmentation and a vertical scrolling operation. The reading speed obtained by the proposed layout of 4.4 characters per line was comparable to the fixed-line length layout of 29 characters per line. This improvement is primarily achieved by a reduction in the number of fixations. Moreover, 85% of the participants did not experience illegibility or incongruousness with the stepwise incremental indent layout reading. These results indicate that this layout is an effective technique to improve the efficiency of reading text with line lengths of 5 characters per line without an increase in cognitive load or a decrease in comprehension. These experimental results indicate that the Japanese electronic text reader with these proposed techniques can improve the reading speed of text with line lengths of 5–40 characters per line without an increase in cognitive load or decrease in comprehension.

    抄録全体を表示
  • 新美 潤一郎, 星野 崇宏
    32 巻 (2017) 2 号 p. B-G63_1-9
    公開日: 2017/03/01
    ジャーナル フリー

    Nowadays, along with the popularity of E-Commerce, the marketing strategy of retail stores has been more complicated with O2O or Omni-channel. Therefore, Customer Relationship Management (CRM) is one of the important issue for the retail stores. It can be difficult to predict customers future behavior with the simple quantitive information such as purchase frequency since each customers are widely diversified. Although the company can obtain the variety of customers information from their online activity, the use of access history is still limited. In this paper, we defined “the variety of user access patterns” collected from their web browsing history and it shows the patterns they visit the website. Finally, we verified its effectiveness with developing a DNN model to predict customers future behavior.

    抄録全体を表示
  • 松井 哲也, 山田 誠二
    32 巻 (2017) 2 号 p. C-G92_1-10
    公開日: 2017/03/01
    ジャーナル フリー

    The PRVAs, product recommendation virtual agent, are the agents that take part in the clerks on the onlineshopping. For their aims, it is very important for the PRVAs to be trusted by users. However, trustworthy the PRVA design was not be studied yet. In this paper, we suggest the user’s trust transition model that is consisted by two parameters. One parameter is user’s emotion, and the other is agent’s knowledge. We suggested the transition operators that transited these two parameters by executing when the PRVAs recommend. Emotion transition operators are agent’s smile and gestures. Knowledge transition operators is long product recommendation text. We carried on three experiments to estimate these model and transition operators. In experiment 1, we executed no transition operators. In experiment 2, we executed emotion transition operators and added knowledge transition operators in the latter half. In experiment 3, we executed knowledge transition operators in the first half and added emotion transition operators in the latter half. As a results, it is discovered that transition operators and a transition model are effective. In experiment 1, there are no transition in the participants’ trust state. In experiment 2, the participants’knowledge perceived and trust for agent transited after executing knowledge transition operators. In experiment 3, the participants’ emotion transited after executing positive emotion operators, however, trust didn’t transited. From these result, we concluded that trust is based on each of the user’s emotion and the agent’s knowledge.

    抄録全体を表示
  • ジメネス フェリックス, 加納 政芳, 吉川 大弘, 古橋 武
    32 巻 (2017) 2 号 p. D-G51_1-12
    公開日: 2017/03/01
    ジャーナル フリー

    This paper sought to examine how behavior of a robot can prompt learning by observing in collaborative learning. The robot learns while solving a problem issued by an English vocabulary learning system with a human learner. The learning system presents English words in example sentences and uses a scaffolding function that helps the learner guess the meaning of English words in the example sentence upon a user request. The robot was designed to solve the questions by using scaffolding function and could not answer correctly at beginning. However, the robot change its question-answering method by guessing the meanings of English words in example sentences and improve its accuracy as learning progressed. This behavior of robot can prompt learners to learn by observing in collaborative learning. Ten college students with low level English learned using the English vocabulary learning system with robot for two months and were videoed during that time to see how they learned. We found that learners learned the English vocabulary by using scaffolding function at beginning. However, learners changed their learning method form using scaffolding function to guessing the meanings of English words in English sentences by learning progress. This suggests that robot, which changes the question-solving method to a more effective one and increases its accuracy rate as learning progress, prompts learners to learning by observing in collaborative learning and change their learning method to the more effective one. This learning by observing indicates that learners learn how to guess the meanings of English words in English sentences by observing the robot’s question-solving and speaking. However, the robot does not prompt some learners to learning by observing because they feel lousy that the robot answers the question and improves its accuracy rate, so they ignore what the robot says. Additionally, learners interest in robot decrease when robot performs the same action.

    抄録全体を表示
  • 増田 壮志, 古崎 晃司
    32 巻 (2017) 2 号 p. E-G71_1-10
    公開日: 2017/03/01
    ジャーナル フリー

    Ontologies are currently constructed in various fields, such as life sciences, medical information, and sustainability science. These ontologies are used as knowledge bases and knowledge models for application systems. However, it is difficult to build high quality ontologies due to the necessity of having both knowledge of ontology and expertise in the target domain. Therefore, ontology construction and maintenance costs considerable time and effort. To reduce such costs, we developed an ontology refinement support method. To test and confirm this refinement method, we focused on the guideline for building well-organized ontologies that“ Each subclass of a super class is distinguished by the values of exactly one attribute of the super class. ”Then, we discovered that there is a similarity between is-a hierarchies when an ontology is built following this guideline and made the hypothesis that, if subclasses are not classified by one attribute, there are consistency errors in the ontology that can be automatically fixed by a comparison method of is-a hierarchies. To test this hypothesis, we conducted an experiment to evaluate the refinement method. We asked nine experienced evaluators to build the ontology and used 150 refinement proposals. As a result, we found that at most 90% of the refinement candidates could be further refined and that at most approximately 50% of the refinement proposals are appropriate to apply to ontologies.

    抄録全体を表示
  • 伊藤 詩乃, 田中 佑岳 , 狩野 芳伸, 榊原 康文
    32 巻 (2017) 2 号 p. F-AI30Ge_1-10
    公開日: 2017/03/01
    ジャーナル フリー

    31 巻6 号AI30-G(2016 年)の論文において、本文引用箇所がすべて[?]として公開されているため、正しい情報を次ページより掲載します.

    抄録全体を表示
    • |<
    • <
    • 1
    • >
    • >|
feedback
Top