Using Reasoning to Decide the Answer to a Decision Question Involving an Opinion Adjective

Communication ability is an important feature of humans. Artificial Intelligence is to simulate human learning, understanding, and thinking processes by a computer program. A Learning Program System (ALPS) has been developed enabling computers to understand knowledge, to learn English grammar, and to read English sentences. Here, we intend to improve the communication ability of ALPS. The focus of this thesis is to understand and answer the decision question involving an opinion adjective. Humans can learn notions easily, like "importance". However, it is difficult for a computer to understand notions well because notions are abstract and difficult to quantify. This thesis analyzes the human's learning process of notions and creates a logical inference algorithm to simulate the human's thinking process. A human learns a new notion by learning the various criteria required to justify the use of an adjective of the notion. The criteria include both examples and important factors. We created a knowledge-based method to simulate the process of learning a notion with various criteria. In our knowledge-based method, we abstract knowledge components of a notion: scales, factors, and examples of each factor. Given the target, the adjective, and the reference that represent a decision question, our inference algorithm first matches the target with the examples. If not matched, it then measures the similarity between the target and the examples based on the reason why each example is considered to be an example. After that, a heuristic search will search the factors. Our program will give a "Yes" if the target can satisfy any factor; otherwise, it will give a "No". In addition, an associated reason is also provided to justify the answer.