| Project/Area Number |
11650412
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
System engineering
|
|---|
| Research Institution | TAKUSHOKU UNIVERSITY |
|---|
Principal Investigator |
ISKIKAWA Tsuyomu Takushoku University Faculty of Engineering Professor, 工学部, 教授 (10281370)
|
|---|
| Project Period (FY) |
1999 – 2001
|
| Project Status |
Completed (Fiscal Year 2001)
|
| Budget Amount *help |
¥3,000,000 (Direct Cost: ¥3,000,000)
Fiscal Year 2001: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 2000: ¥1,400,000 (Direct Cost: ¥1,400,000)
Fiscal Year 1999: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | inference method / Knowledge base of words / predicate logi / common sence / measuring similarity / Dialogue sustem |
|---|
| Research Abstract |
We developed an inference method, called ARSK(Approximate Reasoning with Similar Knowledge), which derives approximate solutions when the knowledge base for solving a problem is incomplete. ARSK is based on the assumption that similar knowledge leads to a similar conclusion. At each inference step, it uses the knowledge similar to the existing knowledge to derive a solution. It also takes the context of the problem into account when measuring the similarity between knowledge.
In order to evaluate ARSK, we developed a dialogue system. This system acquires knowledge about users from input sentences. ARSK is used to derive answers for users' questions based on acquired knowledge. The answers to simple questions obtained by ARSK matched those obtained based on our intuition. Judging from this result, we can say that ARSK is effective in some practical applications.
We also constructed a knowledge base containing approximately 260, 000 Japanese words and 6,000 Chinese characters to judge similarity between knowledge. Each word in this knowledge base is represented by a list of weighted keywords acquired from machine-readable dictionaries. The evaluation results of this knowledge base show that it is adequate in the accuracy of similarity measurement.
|
