Learning about Actions and Events in Shared NeMuS

The categorization process of information from pure data or learned in unsupervised artificial neural networks is still manual, especially in the labeling phase. Such a process is fundamental to knowledge representation [6], especially for symbol-based systems like logic, natural language processing and textual information retrieval. Unfortunately, applying categorization theory in large volumes of data does not lead to good results mainly because there is no generic and systematic way of categorizing such data processed by artificial neural networks and joining investigated conceptual structures. Connectionist approaches are capable of extracting information from artificial neural networks, but categorizing them as symbolic knowledge have been little explored. The obstacle lies on the difficulty to find logical justification from response patterns of these networks [2]. This gets worse when considering inductive learning from dynamic data which is very important to Cognitive Sciences that considers categorization as a mental operation of classifying objects, actions and events [1]. We shall address the discoveries of our on-going investigation on the problem of inductively learning (IL) from dynamic data by applying a novel framework for neural-symbolic representation and reasoning called share Neural Multi-Space (NeMuS) used in the Amao system[4]. Instead of woking like traditional approaches for ILP, e.g. [5], Amao uses a shared NeMuS of a give background knowledge (BK) and uses inverse unification as the generalization mechanism of a set of logically connected expressions from the Herbrand Base (HB) of BK that defines positive examples.