Adaptive and statistical approaches in conceptual modeling
Timo Honkela, Kevin Hynnä, Krista Lagus and Jaakko Särelä
(2005) Technical Report. Publications in Computer and Information Science, Espoo, Finland.

Abstract

Conceptual modeling is a task which has traditionally been conducted manually. In artificial intelligence, knowledge engineers have written descriptions of various domains using formalisms based on predicate logic and other symbolic representations such as semantic networks and rule-based systems. The development of expert systems in 1980s was a notable example of such e orts. As modern, related attempts, the Semantic Web and knowledge representation formalisms like extendable markup language (XML) can be mentioned. Wwe consider the development and application of adaptive and statistical methods for conceptual modeling to be particularly important. Probability theory is an excellent model for dealing with noisy and ambiguous phenomena, such as language. Probabilistic models of linguistic structure exist at every level (phonology, morphology, the lexicon, syntax, discourse). Furthermore psycholinguistic research has shown that probabilities play an important role throughout language comprehension, production and learning. In this publication, adaptive and statistical methods are considered within the area of semantics, in particular. An important aspect is emergence: how representations emerge through a learning or analysis process. Specific topics include emergence of structure (Chapter 2), similarity of emergent representations (Chapter 3), modeling of concepts that are based on multimodal domains (Chapter 4), representation of action (Chapter 5), and category learning (Chapter 6).

[Edit]