Study of database for programs

程序数据库研究

基本信息

批准号：
16500071
负责人：
MIYAMOTO Kenji
金额：
$ 2.08万
依托单位：
Hosei University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2004
资助国家：
日本
起止时间：
2004 至 2007
项目状态：
已结题

项目摘要

The aim of this research is to establish a framework for sharing complex procedural knowledge.Current Semantic Web technology provides a knowledge-sharing framework by sharing meta-level explicit semantics and adding tags indicating their meanings.Semantic Web is supposed, however, for concrete data in specific domain and is not suitable for procedural knowledge such as problem solving.The presented approach emphasizes reusability of structural knowledse composed of many components of rather few kinds. To be properly reused, this kind of knowledge requires additional care of being prepared specifiable, applicable, and validatable.To investigate such mechanism, database of programs was desisned.In storing programs, Floyd-Hoare verification proofs are appended along with their abstract syntax trees.The searchins process descends the tree testins whether each node meets the requirement. If a node which meets the requirement is found, the subtree rooted from it is returned.Receiving the subtree, the client program check the proof tree and instantiate it to concrete syntax with variable names accommodated.As the result, the proposed database meets the above requirements:1. Specifiablility : Logical meaning is used to specify the target program, not keywords or function names.2. Applicability : Although stored in abstract form, retrieved programs are instantiated ready to use.3. Validatability : The meaning of the retrieved program can be certified before actual use.In addition to the above result, it was investigated how the availability is improved.If the first search trial fails, the improved system repeatedly change the requirement so that, if the answer to the new requirement is found, answer to the original one can easily be construct from it.By adding the newly constructed program back to the program store, the database can be fed automatically each time it is used. This self-feeding mechanism makes the database Web2.0.

本研究的目的是建立一个共享复杂过程知识的框架。当前的语义Web技术通过共享元级显式语义并添加指示其含义的标签来提供知识共享框架。然而，语义Web是针对具体数据的所提出的方法强调由许多很少种类的组件组成的结构知识的可重用性。为了正确地重用，这种知识需要额外注意准备可指定的、适用的和可验证的。为了研究这种机制，设计了程序数据库。在存储程序时，附加了Floyd-Hoare验证证明及其抽象语法树搜索过程沿着树向下测试每个节点是否满足要求。如果找到满足要求的节点，则返回以该节点为根的子树。收到子树后，客户端程序检查证明树并将其实例化为包含变量名的具体语法。因此，建议的数据库满足上述要求要求：1。特定性：逻辑意义用于指定目标程序，而不是关键字或函数名称。2．适用性：虽然以抽象形式存储，但检索到的程序已实例化可供使用。3．有效性：在实际使用之前可以验证检索到的程序的含义。除了上述结果之外，还研究了如何提高可用性。如果第一次搜索尝试失败，改进后的系统会反复更改要求，以便，如果找到新需求的答案后，可以轻松地从中构造出原始需求的答案。通过将新构造的程序添加回程序存储，每次使用时都可以自动输入数据库。这种自我反馈机制使得数据库成为Web2.0。