In this paper, we overview meaning, definition, and applications of ontology on Artificial Intelligence field. Ontology on AI stems from knowledge representation, in particular, knowledge sharing problems, because the traditional ways of knowledge representation lacks completeness, exhaustiveness, and systematicness for sharing and re-use of knowledge. Ontology is therefore expected to be a base of knowledge sharing and reuse. Ontologies have some types, i.e., top-level, domain, task, and application ontologies, and some levels from one providing names and relations to one providing a model for behavior. We also show typical applications of ontologies, i.e., use of ontologies for inter-agent, human-system, and inter-human communication.

View full abstract

　To substantially promote the integration of heterogeneous knowledge and to further facilitate the discovery of fundamental concepts or hidden structures in biological phenomena, we study genome ontology where biological knowledge is comprehensively reorganized and both the terminology and the concepts are unified across species. We try to establish such ontology of signal transduction and to recompile a dictionary of terms in the universe of living systems.

View full abstract

　This paper reports an evaluation method for Japanese Kanji input system based on frequency information and user's input history information. Japanese Kanji Frequency information follows Zipf's law, user's input history information is follow LRU algorithm's hit rate, reports the result conducted evalutaion experiment of this. And considere about a future Japanese Kanji input system.

View full abstract

　The Japanese character set consists of Hiragana, Katakana, and Kanji. Some Kanji characters have the variants that have the same meaning and pronunciations. The users of full-text retrieval systems sometimes confuse some Kanji characters with the variants and get insufficient result for their needs. The authors considered that a function that unifies Kanji variants is need for full-text retrieval systems. This paper describes a full-text retrieval system that has functions of unifying Kanji variants. The system has a thesaurus of Japanese Kanji variants and uses it for indexing and unifying Kanji characters. The authors built two example retrieval systems based on the system. They have the database of ULIS-DL (http : //lib.ulis.ac.jp/) metadata and Japan-MARC.

View full abstract

　To facilitate retrieving patent information across languages, we are developing a multi-1ingual patent retrieval system, where user queries are translated into the target language by way of a dictionary. In this paper, aiming to enhance the translation dictionary, we propose a method to automatically extract translations from Japan-US patent families consisting of Japanese/English comparable texts. Our method computes the association score for each combination of Japanese/English words in patent families, and selects those with greater scores as translations. We also show the effectiveness of our method through experiments.

View full abstract

　It has been three years since the specification of XML was announced. In the meantime, recommendations have been also made for technologies for applying XML to applications. Here we summarize the merit of XML technology and introduce the application to an actual data system, i.e., databases of WFCC-MIRCEN World Data Centre for Microorganisms (WDCM).

View full abstract

　This paper describes the environment that supports selection of Z39.50 databases for users. The environment allows users to share information of Z39.50 databases in WWW. The present authors develop the WWW-Z39.50 client. In this environment, the users can share information which other users organize by themselves. When a user clicks one of list of Z39.50 databases, which is shared information, the window for retrieval of the database will be opened for the user to start his her retrieval immediately.

View full abstract

　Z39.50 is the international standard of an information-retrieval protocol, and can offer the history-reference search function, the environment independent from a specific vender, the crossing reference function with unified interface, etc. In the result of the questionnaire and interview to the university library which is the user of Z39.50 and is an information addresser, it is observed as choice of the technical element in the information disclosure, in spite of the very low domestic popularization level. Although many problems remain in the present condition, it is important not to judge Z39.50 from the viewpoint of only the advantage in the present condition, but to judge on the basis of the future possibility.

View full abstract

　The definition of the "digital archives" is ambiguous; various types of organizations, such as web archives, digital museum, data archives and so on, which collect original material in digital form has been called digital archives. This study gives some concrete examples of digital archives and discusses their good effect and problems.
　Although the good effect given by digitalizing materials is acknowledged widely, the evaluation of digital archives is not easy; for instance, the cost effectiveness is difficult to estimate. One of the roles of the digital archives is to preserve cultural assets; most of them have been founded and operated by government or public organizations. The future of the digital archives might depend on whether the value of them would be given social consensus of their cost effectiveness.

View full abstract

　In recent years, electronic commercial transactions have been getting popular. We attend to develop an electronic shopping mall available to the Internaet. This shopping mall is constructed by a server computer for receiving the orders together with existing databases of the stores which are connected with the server computer. For giving fresh information of the goods to the purchasers, it is indispensable that the server computer and each trader's database frequently exchange messages. We also supply the methods which support the communication.
　Gathering traders of the same goods will lead to a commercial competition. As a result, purchasers will buy fine goods at a lower cost, while the sales dealer will make it easier to secure the new customers and repeaters.

View full abstract

In this paper I described the Problems of the Contents Distribution System in the Commercial Publishing Business in Japan. Last Time I reported the Influence of the Internet on the Commercial Publishing Business. This time I discussed the changes of the environment of the contents distribution system in the commercial publishing. What was changed in the viewing or browsing software and in the charging mechanism of Internet contents? And also I described some new trials in the world.

View full abstract

　We have been making some attempts for a new paradigm of Information and Knowledge Science. Researching our specific subject "Database for getting new ideas or thoughts", we are always feeling some intellectual gap between the Western thinking manners and ours. Naturally, present Information and Knowledge Science is established on Western Civilizations. On the other hand, our thinking manner is based on Japan Culture. Recently they say that our Japan Culture is unique among many World Civilizations. So we hope new frameworks of which foundation is constructed the complexes of Japan Culture and others. Toward this difficult but pleasant subject, we began to make a small first steep. At present we are interesting in the following ideas as the viewpoints.
　1. Our hope is that the New Paradigms must be based on not only Western Civilizations but also Eastern and Japan Culture.
　2. We have many types of unique logic, arts and skills in our culture. Through the Ancient and Middle times, some of them were born in Japan islands and others came from Korea, China and India
　3. Now we are trying to discover new merits among them from the viewpoint of information system.

View full abstract

　SS-SANS method is to extract semantic relationships with automatic. This work is to apply SS-SANS extract in a associative relationships in the test collection which includes thesis data(300 thousands of japanese abstracts). The process uses templates between two terms for extraction and it extracted automaticaly 150 thousands of relationships. These relationships find out distinction between front term and end term. Front terms are called OBJ terms and end terms are called SBJ terms. Analyse of terms for OBJ terms or SBJ terms and their distributions. In conclusion, distinctions to use terms for OBJ terms or SBJ terms are explained.

View full abstract

　Semantic processing which uses a great deal of knowledge which was structuralized require to be equipped with learning and consideration functions in computers. The conceptual structure according to conceptual relation exists as a construction of structuralized information that is necessary to semantic processing. Systematizing terminology as minimum unit which expresses a concept is most important to semantic processing. It made conceptual structure which combined the C-TRAN method extracting equivalent relationships and the SS-KWEIC method which extracts hierarchy relationships and associative relationships, moreover extracted data by the SS-SANS method extracting associative relationships. Hence the new relationships are found among terminologies and judgment material for semantics of terminology increases. Finally it became possible to get more detailed terminologies to systematize.

View full abstract

　The global flow of information is being developed at unprecedented speed, and an importance of information retrieval become higher. However, users may not make a good use of huge amount of information by using conventional computers whose major functions are numerical calculation, symbol matching in information retrieval and deduction. Especially, an importance of information retrieval becomes higher. However, it is difficult to search relevant information that is satisfied its purpose from a vast information efficiently.
　To solve these problems, a new intelligent method of information retrieval using organized knowledge resources based on semantic relationships is proposed.

View full abstract

　This work is a report on the use of a language application using semantic representation in an XML format. The aim of the application is to correct the misuse of Japanese postpositional words, called joshi as they have been used in sentences written by non-Japanese students. First, the sentences written by the students are processed morphologically so that pairs of joshi nouns and pairs of joshi verbs are extracted. Next, document data that include the same extracted nouns are picked up from an EDR corpus where these nouns are stored with other pairs of joshi that are used as training examples in the correction process. The data are automatically transformed into semantic representations in an XML form. "Association rules" are applied in the correction process to assist in learning how to use the appropriate joshi after target nouns. Semantic representation is used to provide suitable explanations of how joshi should be used in a sentence.

View full abstract

　In this study, relationships between the Information Science and other fields were investigated in terms of keywords extracted from research themes of each researcher. First, procedure for extracting keywords from the database of Directory of Researchers was presented. As an index measuring the relationship, we introduced a coefficient of normalized complete congruence and further applied the Correspondence Analysis method to analyze the relations between the three items of the so-called Information Science and other related research fields at Japanese universities.

View full abstract

　In the Net News, conflicts among the participants are often observed. It seems that the conflicts are often caused by the difference in the cultural background of the participants. In this paper, we intend to propose a quantitative extraction method of the conflicts. Articles in the Net News can be positioned in tree structures (threads) defined by the referential relation. A thread is treated as a set of samples for the analysis. Each article is distinguished by key word frequencies and they are summarized to principal component scores. Applying this technique to actual Net News threads, we found that when a thread includes conflicts, some scores vibrate with wavelength of two generations in the thread. Also the key of the conflict can be estimated with the meaning of the vibrating principal component. We expect that such a quantitative extraction technique of conflicts is useful for study of the social relation through electronic communication.

View full abstract

[in Japanese]

View full abstract