Iconic language design for people with significant speech and multiple impairments
P. L. Albacete, S.-K. Chang, G. Polese, B. Baker · 1994 · Proceedings of the First Annual ACM Conference on Assistive Technologies (Assets '94) · doi:10.1145/191028.191035
Summary
This paper presents a formal methodology for designing iconic languages used in augmentative and alternative communication (AAC) systems for people with significant speech and multiple impairments (SSMI). The authors build on the Minspeak system, conceived by co-author Bruce Baker, which uses the principle of semantic compaction — mapping concepts onto multi-meaning icon sequences that are decoded by a microcomputer with speech synthesis output. With only 50 to 120 icons on a keyboard, the system can represent around two thousand concepts by combining ambiguous icons into iconic sentences whose meaning is resolved by the sequence. The paper formalizes this design process using two theoretical frameworks: Icon Algebra, which provides formal operators (combination, marking, context, enhancement, inversion) for deriving new meanings from icon pairs, and Conceptual Dependency theory, adapted from natural language processing to serve as a semantic model for iconic sentences. The authors replace CDT's primitive actions with primitive connectors that correspond to each Icon Algebra operator, creating CD forms that systematically capture possible meanings of iconic sentences. An inference engine, implemented in C, operates in both specialized and exhaustive modes to generate candidate meanings for icon combinations, which the designer can then accept, modify, or discard.
Key findings
The paper demonstrates that the semantics of iconic sentences can be formalized and computationally derived, moving icon-based AAC language design from an ad hoc process to a principled methodology. The inference engine successfully generates possible meanings for icon combinations by applying CD forms associated with each Icon Algebra operator, selecting relevant frame attributes from individual icons and composing them according to operator semantics. The system helps designers choose the most intuitive iconic sentence to encode a given word by systematically browsing all possible meanings, which may improve consistency across multiple designers working on the same language. The authors note that the Minspeak system had already been distributed in over 20,000 units worldwide by 1994, with Swedish, German, and Italian versions being developed, demonstrating the real-world viability of semantic compaction as an AAC strategy. The approach compares favorably to Blissymbols (which burden the listener with inference) and abbreviation systems (which become unwieldy at scale).
Relevance
This paper addresses a fundamental challenge in AAC design that persists today: how to enable rich, expressive communication using a small set of symbols accessible to people with significant motor and cognitive impairments. The Minspeak system and semantic compaction approach described here became one of the most commercially successful AAC paradigms, with Prentke Romich Company devices still widely used. The formal design methodology is significant because it provides a principled way to evaluate and optimize iconic language designs for learnability and intuitiveness — key factors in AAC adoption and abandonment rates. The work anticipates modern concerns about AAC vocabulary design, where the balance between expressive power, learning difficulty, and communication speed remains a central tension. The integration of computational inference tools into the language design process also foreshadows current interest in AI-assisted AAC, where machine learning helps predict and generate communications from limited user input.
Tags: AAC · iconic language · semantic compaction · Minspeak · speech impairment · multiple impairments · augmentative communication · symbol-based communication · icon algebra · conceptual dependency · visual language · Blissymbols · language design