Published papers

Kana-kanji conversion system is widely used in Japan. It is convenient for users that other systems use a metaphor of kana-kanji conversion in order to ease the learning cost. Regular expression generation tool is implemented to show usefulness of the metaphor. First, the user gives a sample sentence, from which the tool generates a candidate of regular expression. The candidate is a concatenation of frequently used expressions stored in the dictionary. Next, the user modifies the candidate, where operations are very similar to those of egg, a kana-kanji conversion interface for emacs. The metaphor of kana-kanji conversion is limited to dealing with sequential concatenation of expressions in the dictionary. To generate hierarchical structure, cascading multiple layers is proposed. Each layer generates a sequential structure with the metaphor, and the result sequence is sent to the upper layer as an input string. In upper layers, the dictionaries consist of rewriting rule for generating complex structures.

The executable constraint is proposed to offer designers a way to define a new set of constraints on top of an editing tool. It aims mainly to maintain the integrity of data structures such as diagrams that users create, during an edit session. When a part of the structure is modified by a user, the computer modifies other parts as a consequence of the user's instructions given in advance. Executable constraints are sequences of editing operations whose fixed points give relations to be maintained. They are constructed from primitive editing operations by demonstration to avoid textual form of expression, so that non-programmers can define new relations. Priority and condition are introduced to cope with over-constrained or conditional situations.

An operation sequence is the basic structure used in the programming by demonstration. It does not, however, provide any sophisticated abstraction mechanism that is necessary to maintain large-scale programs. This paper proposes the use of divided operation sequences instead of the single one. By using multiple sequences, programmers can focus on one of the sequences, that consists of only interested events. Divided operation sequences may be useful for describing and debugging complicated programs.

Location information is important to construct context-aware applications. A significant location is a specialized form of location context for expressing a user's daily activity. We propose a method to learn significant locations from positions measured by cellular phones. Cellular phones we used are equipped with a positioning system, where data can be taken in low frequency with wide-varying estimated errors. In order to learn significant locations, our system uses a clustreing method which exploits multiple radii for coping with these characteristics and for adapting to a variety of users' spatial behavioral patterns. We also discuss characteristics and appropriate parameters for our method.

多面分類法とUMLを用いた、オブジェクト指向を用いて作成されたソフトウェアの再利用時における、再利用ソフトウェアライブラリからの検索方式について提案する。特にオブジェクト間のデザインパターンのような関係を多面分類を用いて分類し、的確に再利用可能なソフトウェアを検索する。これを用いる事でオブジェクト指向を用いて作成されたソフトウェアの再利用において、オブジェクト間のパターン的な関係を検索出来るという点で検索が効果的に行なえるようになる。

The executable constraint aims to maintain the integrity of structures that users create during an edit session. When the users modify a part of the structures, other parts are modified accordingly, based on instructions given by the users. The instructions are presented by demonstration, and form a constraint that is satisfied when the current state is its fixed point. The users can dynamically control a set of executable constraints to be satisfied.

Litesに対する並列分散処理機能の拡張の一部として、並列分散データストリーム処理機能を提供する。従来のUNIXにおけるパイプ、ソケットなどのデータストリームの概念を拡張した形で並列化・分散化を行い、既存プログラムが大きな変更なしに並列分散環境においてデータストリーム機能を利用することを可能とする。

Litesに対する並列分散ファイル入出力機能拡張について解説する。本機能はUNIXのファイルの概念を拡張し，複数のプロセスが並列的にファイルアクセスすることを可能とする機能である。本機能は単体でも機能するが，並列拡張されたlites上で提供される並列分散プロセス管理や並列分散パイプ機能と組み合わせることによってより効果的に機能する。

並列分散プログラム開発・実行環境を提供するためLitesの拡張を行う。本サーバはマイクロカーネル上で動作し，並列分散プロセスに対してプロセス管理や共有記憶，データストリーム，ファイル入出力などのサービスを行う。これらのサービスはUNIXのパイプやファイルを拡張する形で実現され，従来のプログラミングスタイルを保持したまま並列分散プログラムに必要な機能を提供する。