跳到主要内容

東京大学 情報理工学系研究科 創造情報学専攻 2008年8月実施 筆記試験 第3問

标签：

Author​

itsuitsuki

Description​

There are several methods for knowledge representation. In contrast to a system of production rules and logical formulae which are modular representations, the semantic network (where a node represents a concept or an entity and a named link represents a relation between the nodes) has the characteristics of representing the pieces of knowledge mutually interrelated. For example, the following knowledge can be represented in a semantic network shown on the right.

  • A pigeon is a bird.
  • A bird has wings.
  • A pigeon eats soybeans.

Q1. Represent the following pieces of knowledge 1)∼8)1)\sim8) as a semantic network. (Here, you can create necessary link names. In this case, use a meaningful name for a link, or append an explanatory comment so that the meaning of the link name can be understood.)

  1. Taro is a human, and works in the Ueno Zoo.
  2. Taro takes care of Ueno Zoo's penguins.
  3. Yuki is one of Ueno Zoo's penguins, and lives in the Ueno Zoo.
  4. A penguin is a bird.
  5. A bird can fly.
  6. Taro owns a Prius.
  7. Kenta is a human, and owns another Prius.
  8. Prius is a hybrid car.

Q2. (2-1) In the semantic network made in Q1, "Yuki can fly" can be derived. Explain the reason why this derivation occurs. (2-2) Describe which widely used programming languages include functionality similar to what occurs in (2-1). (2-3) Explain the merits of the functionality mentioned in (2-1) and (2-2) from the viewpoint of the description and management of knowledge or programs.

Q3. We want to efficiently infer an answer to the query “who is the human that takes care of Yuki and owns a hybrid car?” in the semantic network made in Q1. Explain what type of inference method can be employed, and find an answer for this query.

Q4. Assume that the following knowledge is added to the semantic network made in Q1.

  1. A penguin cannot fly.

In this case, there are the following possibilities.

a) Both “Yuki can fly” and “Yuki cannot fly” can be derived. b) Neither “Yuki can fly” nor “Yuki cannot fly” can be derived. c) Only “Yuki cannot fly” can be derived.

Why do such possibilities arise? Explain what type of knowledge manipulation is required to yield the case c).

题目描述​

知识表示有多种方法。产生式规则和逻辑公式通常以模块化方式表示知识；语义网络则以结点表示概念或实体、以带名称的边表示结点间关系，能够显式表现知识片段之间的联系。原文示例把“鸽子是鸟”“鸟有翅膀”“鸽子吃大豆”表示成语义网络。

  1. 将以下 8 条知识表示成语义网络。可以自行创建关系边名称，但名称应语义明确；否则须附注说明其含义。
    1. 太郎是人，并在上野动物园工作。
    2. 太郎照料上野动物园的企鹅。
    3. Yuki 是上野动物园的一只企鹅，并住在上野动物园。
    4. 企鹅是鸟。
    5. 鸟会飞。
    6. 太郎拥有一辆 Prius。
    7. 健太是人，并拥有另一辆 Prius。
    8. Prius 是混合动力汽车。
  2. 回答有关第 1 问语义网络的继承机制：
    1. 说明为什么能够推出“Yuki 会飞”。
    2. 指出哪些广泛使用的编程语言具有与上述现象相似的功能。
    3. 从知识或程序的描述与管理角度，说明这种功能的优点。
  3. 要高效回答“谁是照料 Yuki 且拥有混合动力汽车的人？”这一查询，说明可采用何种推理方法，并给出查询答案。
  4. 再加入知识“企鹅不会飞”。此时可能出现三种处理结果：a. 同时推出“Yuki 会飞”和“Yuki 不会飞”；b. 两者均不能推出；c. 只推出“Yuki 不会飞”。说明为何会有这些可能性，以及要得到结果 c 需要怎样的知识处理。
本页目录