Title: Acquiring Design Problem Reformulation Knowledge: On computable relationships between design syntax and semantics
Speaker: Somwrita Sarkar
Affiliation: School of Architecture, Univ. of Sydney (Ph.D. Candidate)
Date: February 3, 2009
Abstract:
The relationship between the syntax of design representation and the encoded semantic meaning is an open research question. There are no existing cognitive or computational models on how ill-defined, ambiguous knowledge transforms into formal, well-structured, mathematical-symbolic representations. Design modeling and reformulation tasks exemplify this learning problem. Existing research shows that the largely prevalent symbolic AI ''rule based'' paradigm leads to high levels of knowledge engineering and large training databases in design learning and automation systems.
This seminar presents a computational method employing an unsupervised, knowledge-lean, pattern recognition and extraction approach for the inductive inference of explicit and implicit semantic design knowledge from the symbolic-mathematical syntax of design formulations to automate a range of problem reformulation tasks.
Findings suggest that, in formal design representations, the seat of semantic meaning may lie as much in the implicit (global) patterns of relationships between symbols, inferred from their explicit (local) contextual co-occurrence patterns, as it lies in the explicit local mapping of a symbol to its semantic meaning. This understanding can form an alternate basis for developing learning algorithms for design computation tasks like problem reformulation.
Brief bio of speaker:
Somwrita Sarkar has a B.Arch. in Planning from Delhi School of Planning and Architecture, and is currently finishing up her PhD at the Key Center for Design at Univ. of Sydney.