Evaluating pictogram prediction in a location-aware augmentative and alternative communication system

Abstract

This study compared the performance of two statistical location-aware pictogram prediction mechanisms, with an all-purpose (All) pictogram prediction mechanism, having no location knowledge. The All approach had a unique language model under all locations. One of the location-aware alternatives, the location-specific (Spec) approach, made use of specific language models for pictogram prediction in each location of interest. The other location-aware approach resulted from combining the Spec and the All approaches, and was designated the mixed approach (Mix). In this approach, the language models acquired knowledge from all locations, but a higher relevance was assigned to the vocabulary from the associated location. Results from simulations showed that the Mix and Spec approaches could only outperform the baseline in a statistically significant way if pictogram users reuse more than 50% and 75% of their sentences, respectively. Under low sentence reuse conditions there were no statistically significant differences between the location-aware approaches and the All approach. Under these conditions, the Mix approach performed better than the Spec approach in a statistically significant way.

Pictogram Prediction in Alternative Communication Boards: a Mapping Study

“…For this reason, we opted to maintain their label classification separated instead of grouping them as knowledge bases. Two studies using statistical language models based on n-grams [Hervás et al 2020, Garcia et al 2016. These studies trained bi-gram language models by using pre-defined text corpora.…”

“…They tested how well the controlled language (i.e., set of sentences) fits into the semantic grammar. [Garcia et al 2016] ran several software simulations to measure the performance of the different pictogram prediction approaches they proposed. They evaluated the models over a set of sentences indicated by specialists as adequate for the AAC domain.…”

“…[Pereira et al 2020] evaluated their proposal's precision for reconstructing the sentences from a corpus. And finally, [Garcia et al 2016] assessed the system's quality on saving keystrokes. This way, there is not a consensus on what metric is most adequate to the task.…”

“…As discussed in Section 1, in high-tech AAC systems, each pictogram has an associated label or caption, which can be a word or a multi-word expression. Some of the analyzed studies consider this label enough for making pictogram prediction [García et al 2015, Saturno et al 2015, Garcia et al 2016, Hervás et al 2020. This way, they perform a word prediction and do not take care of polysemic words.…”

Predictive composition of pictogram messages for users with autism

“…In (Garcia et al 2016) the authors evaluated the performance of Eugénio, an AAC system for Portuguese, by measuring the effects of adding location knowledge to its pictogram prediction system. The location-aware pictogram prediction method was trained using a corpus composed by able-bodied persons in three different locations: classroom, cafeteria and home.…”