Dialogue in Hierarchical Learning of a Concept Using Prototypes and Counterexamples
نویسندگان
چکیده
While dealing with vague concepts often it puts us in fix to determine whether to a particular situation/case/state a particular concept applies or not. A human perceiver can determine some cases as the positive instances of the concept, and some as the negative instances of the same; but there always remain cases, which might have some similarities with some positive cases, and also have some similarities with some negative cases of the concept. So we propose to learn about the applicability of a concept to a particular situation using a notion of similarity of the situation with the available prototypes (positive instances) and counterexamples (negative instances) of the concept. Perceiving a vague concept, due to the inherent nature of vagueness, is subjective, and thus never can be exhausted by listing down all the positive and negative instances of the concept. Rather we may come to realize about the applicability, or non-applicability, or applicability to some extent, of a concept to a situation in a step-by-step hierarchical manner by initiating dialogue between a perceiver and the situation descriptor. Hence, the main key ingredients of this proposal are (i) prototypes and counterexamples of a concept, (ii) similarity based arguments in favour and against of applicability of a concept at a particular situation, and (iii) hierarchical learning of the concept through dialogues. Similarity based reasoning [3], hierarchical learning of concepts [1], dialogue in the context of approximation space [2] all are separately important directions of research. For our purpose, in this presentation we would concentrate on combining these aspects from a different angle. In [4], a preliminary version of logic of prototypes and counterexamples has been set. To make this paper self-contained, we recapitulate the necessary definitions below. We start with a set S of finitely many situations, member of W may be called a world. We now consider a fuzzy approximation space W, Sim, where Sim is a fuzzy similarity relation between worlds of W. That is, Sim : W × W → [0, 1], and we assume Sim to satisfy the following properties. (i) Sim(ω, ω) = 1 (reflexivity) (ii) Sim(ω, ω ′) = Sim(ω ′ , ω) (symmetry) (iii) Sim(ω, ω ′) * Sim(ω ′ , ω ′′) ≤ Sim(ω, ω ′′) (transitivity). Following [3], the fuzzy approximation space W, Sim is based on the unit
منابع مشابه
On-Line Learning of a Persian Spoken Dialogue System Using Real Training Data
The first spoken dialogue system developed for the Persian language is introduced. This is a ticket reservation system with Persian ASR and NLU modules. The focus of the paper is on learning the dialogue management module. In this work, real on-line training data are used during the learning process. For on-line learning, the effect of the variations of discount factor (g) on the learning speed...
متن کاملOn-Line Learning of a Persian Spoken Dialogue System Using Real Training Data
The first spoken dialogue system developed for the Persian language is introduced. This is a ticket reservation system with Persian ASR and NLU modules. The focus of the paper is on learning the dialogue management module. In this work, real on-line training data are used during the learning process. For on-line learning, the effect of the variations of discount factor (g) on the learning speed...
متن کاملHierarchical Functional Concepts for Knowledge Transfer among Reinforcement Learning Agents
This article introduces the notions of functional space and concept as a way of knowledge representation and abstraction for Reinforcement Learning agents. These definitions are used as a tool of knowledge transfer among agents. The agents are assumed to be heterogeneous; they have different state spaces but share a same dynamic, reward and action space. In other words, the agents are assumed t...
متن کاملHigh-Dimensional Unsupervised Active Learning Method
In this work, a hierarchical ensemble of projected clustering algorithm for high-dimensional data is proposed. The basic concept of the algorithm is based on the active learning method (ALM) which is a fuzzy learning scheme, inspired by some behavioral features of human brain functionality. High-dimensional unsupervised active learning method (HUALM) is a clustering algorithm which blurs the da...
متن کاملHierarchical Reinforcement Learning for Spoken Dialogue Systems
This thesis focuses on the problem of scalable optimization of dialogue behaviour in speech-based conversational systems using reinforcement learning. Most previous investigations in dialogue strategy learning have proposed flat reinforcement learning methods, which are more suitable for small-scale spoken dialogue systems. This research formulates the problem in terms of Semi-Markov Decision P...
متن کامل