Improving Word-Finding in Assistive Communication Tools: A Mixed-Initiative Approach (thesis)

Abstract:

Navigating a vocabulary consisting of thousands of entries in order to select appropriate words for building communication is challenging for individuals with lexical access disorders like those caused by aphasia. Most existing assistive communication vocabularies have a lexical organization scheme based on a simple list of words. Some word collections are organized in hierarchies which often leads to deep and confusing searches; others are simply a list of arbitrary categories which causes excessive scrolling and a sense of disorganization. Ineffective vocabulary organization and navigation hurt the usability and adoption of assistive communication tools and ultimately fail to help users build functional communication.

We argue that to provide effective word-finding, an assistive vocabulary needs to adapt to individual user’s word usage patterns and to the semantic associations present in a speaker’s mental lexicon, where words are stored and organized in ways that allow efficient access and retrieval. To test our thesis, we employ a mixed-initiative approach to the design of the Visual Vocabulary for Aphasia (ViVA). ViVA is adaptive in that it automatically models an individual user’s mental lexicon according to psycholinguistic theories that propose a semantic network structure of the lexicon and spreading activation as supported by semantic priming. The vocabulary is also adaptable and can be customized to reflect user preferences. ViVA compensates for some of the impaired links in a user’s mental lexicon by building a dynamic network where words are linked based on semantic association measures, human judgments of semantic similarity and past vocabulary usage. Thus, the tool tailors the vocabulary organization according to both user-specific information and general knowledge of human semantic memory.

We evaluated how our system performs compared to a widely used vocabulary access system in which words are organized hierarchically into common categories and subcategories. The results indicate that word retrieval is significantly better with ViVA. In addition, we present results from a longitudinal single-case study with an aphasic participant which illustrates the importance of personal associations for creating an effective assistive vocabulary such as ViVA.