Yasuhiko Kitamura

Language Grid Association: Action Research on Supporting the Multicultural Society

The Language Grid is a middleware with which people can connect and use language resources such as machine translations, morphological analyzers and others created in the fields of intercultural collaboration. The Language Grid cannot exist without the collaboration of Language Grid Users who provide language and computation resources, language services, and collaboration tools. This paper overviews Language Grid Association, a user group of the Language Grid and a body promoting action research to support the multicultural society.

Smoothing human-robot speech interactions by using a blinking-light as subtle expression

Speech overlaps, undesired collisions of utterances between systems and users, harm smooth communication and degrade the usability of systems. We propose a method to enable smooth speech interactions between a user and a robot, which enables subtle expressions by the robot in the form of a blinking LED attached to its chest. In concrete terms, we show that, by blinking an LED from the end of the users speech until the robots speech, the number of undesirable repetitions, which are responsible for speech overlaps, decreases, while that of desirable repetitions increases. In experiments, participants played a last-and-first game with the robot. The experimental results suggest that the blinking-light can prevent speech overlaps between a user and a robot, speed up dialogues, and improve users impressions.

Interactive Integration of Information Agents on the Web

World Wide Web contains a vast amount of different information stored in a huge number of distributed Web sites. Search engines and information agents have been developed to facilitate efficient information retrieval tasks from the Web. By integrating multiple search engines and information agents as an interoperable system, we increase the value of each of them. In conventional collaborative systems, the integration process is designed by system designers and is concealed from the end users. This paper proposes an interactive multiagent-based interface called Multiple Character-agent Interface (MCI) where animated character-agents interact with each other and with the user for assisting in information retrieval. By using the MCI, even a novice user can create a team of information agents and can self-customize the agents through the interactions with them. We here report the architecture of MCI and two prototype systems based on MCI, Venus and Mars, which is a cooperative multiagent system for information retrieval, and Recommendation Battlers, which is a competitive multiagent system for information recommendation.

A Competitive Information Recommendation System and Its Behavior

Information recommendation systems draw attention of practitioners in B-to-C electronic commerce. In an independent recommendation system such as in www.amazon.com, a user cannot compare the recommended item with ones from other information sources. In a broker-mediated recommendation system such as in www.dealtime.com, the broker takes the initiative of recommendation, and the information provider cannot recommend its item directly to the user.In this paper, we propose a competitive information recommendation system consisting of multiple animated agents that recommend their items competitively, and discuss the advantages through showing a prototype developed for restaurant recommendation. Each agent recommends restaurants from its own point of view and the user tells good or bad about them. In our competitive information recommendation system, the user can compare items recommended from multiple agents, and the information providers can recommend their items directly to the user through its animated agent. We also show that the competitive nature affects the output depending on the number of participating agents.

Embodied conversational agent based on semantic web

Embodied conversational agents (ECA’s) are cartoon-like characters which interact with users through conversation and gestures on a computer screen. ECA makes human computer interactions more friendly because we can use most human-like communication skills such as natural conversation. ECA’s are useful as Web guides by incorporating them into Web browsers. They guide us around Web pages chatting with us. To build such an agent, we need to describe a scenario to explain Web pages. Conventionally such scenarios are written manually by developers or programmers using a dialogue description language such as AIML (Artificial Intelligence Markup Language), so it is difficult to update them when Web pages are updated. In this paper, we propose a scheme to automatically generate utterances of Web guide agents depending on Web pages. To this end, we need to make agents understand the contents of Web pages and to make them talk according to the contents, so we utilize RDF (Resource Description Framework) to present the semantic contents of Web pages. To make agents talk according to the contents, we utilize a RDF query language SPARQL (Simple Protocol And RDF Query Language) and extend the AIML language to incorporate SPARQL query in it. As a prototype, we developed a Web guide system employing an ECA.

Multiple character-agents interface: an information integration platform where multiple agents and human user collaborate

We propose the Multiple Character-Agents Interface (MCI) as an information integration platform where multiple animated life-like characters interact with each other and with the user to retrieve and integrate information from the Internet. The MCI makes the process open to the user and allows him/her to collaborate with the character-agents. We implemented the MCI as a multi-agent system in which information agents distributed over the Internet are integrated, then we developed a prototype called Venus and Mars, which is a cooperative cooking recipe search engine consisting of three character-agents that collaborate with the user to locate cooking recipe pages.

Semantic Wiki Where Human and Agents Collaborate

A Wiki is a collaborative Web page authoring system. Users collaborate to build a Web site by creating and updating Wiki pages through Web browsers. However, conventional Wikis easily lose the consistency of the contents because a number of anonymous users can participate in authoring them. By introducing information agents that understand the. Wiki contents, we can keep the consistency. The agents can automatically update Wiki contents, integrate other Web contents to them, and keep them consistent cooperating with the human users. We propose KawaWiki, which is a semantic Wiki system where human users and information agents can collaborate by utilizing the semantic Web technology. To make agents and users collaborate in authoring Wiki contents, we adopt the RDF as the common representation. It is not easy for novice users to author RDF data, and we introduce KawaWiki templates to generate a Wiki page with RDF data at one time. We also introduce KawaWiki queries to make agents retrieve information efficiently from the Wiki contents. Finally, we introduce an agent description language to specify agents behavior on the Wiki.

Action Sloping as a Way for Users to Notice a Robot's Function

This paper focuses on the problem that will arise in the near future from multi-function robots. Users will have to read thick operation manuals to use them. If users can use these robots without reading difficult manuals, it will improve user efficiency. We then proposed action sloping as a way for users to naturally recognize a robots function. It provides the robots with gradual feedback signals when the user performs given actions. By changing the intensity of the feedback signal according to his/her action, it encourages him/her to perform an action that will trigger the robots function. In our experiments, we made three kinds of feedback behaviors according to Action Sloping and one non-feedback behavior as the control condition. The participants of the experiment tried to find a robots function and the latencies to first finding the triggered action were measured. An analysis of the latencies show ed the difference between the sound feedback group by action sloping and the control group. This result showed that the effectiveness of action sloping was partially supported.

Single-agent and Multi-agent Approaches to WWW Information Integration

The WWW is a most popular service on the Internet and a huge number of WWW information sources are available. Conventionally we access WWW information sources one by one by using a browser, but WWW information integration gives a unified view to users by integrating multiple WWW information sources elaborately. In this paper, we introduce our single-agent and multi-agent approaches to WWW information integration.

Multi-state commitment search

We propose the multi-state commitment (MSC) method to speed-up heuristic search algorithms for semi-optimal solutions. The real-time A* (RTA*) and the weighted A* (WA*) are representative heuristic search algorithms for semi-optimal solutions and can be viewed as single-state and an all-state commitment search algorithms respectively. In these algorithms, there is a tradeoff between the risk of making wrong choices in search process and the amount of memory for the recovery, with RTA* and WA* being the extremes. The MSC method introduces a moderate and flexible characteristic into these algorithms and can increase the performance dramatically in problems such as the N-puzzle. In this paper, by introducing a commitment-list, we show a modification of RTA* and WA* to their MSC versions without violating their completeness. Then, we experiment with their performance in maze and N-puzzle problems, and discuss conditions that the MSC method is effective.