Objective:

Describe an application to assist in the speech rehabilitation of participants with verbal apraxia, listing phonemes by degree of difficulty of execution (in ascending order)

Background:

Verbal apraxia involves functional learning: the more interaction there is with speech production itself, the more skilled the speaker will become. However, the ratio of speech therapists to patients is 1:10,000 in the state of Piauí, resulting in a gap in speech therapy services.

Design/Methods:

Four speech therapists developed a network of stimuli for the rehabilitation of acquired neurological disorders, with different levels of complexity (articulation, phonemes, automatic and spontaneous speech, repetition, among others). The network was modeled using Artificial Intelligence and was designed for adults with minimal literacy and preserved auditory and visual functions, using written, acoustic, and visual stimuli. The research included three groups: 1)healthy individuals; 2)speech therapists; and 3)people with speech apraxia. The study was approved by the IFPI Research Ethics Committee (Opinion No. 5111083), respecting ethical precepts and the Brazilian LGPD.

Results:

The study used the FalaBrasil language model as a basis, developed by the Signal Processing Laboratory at UFPA. The stimulus network, modeled with Artificial Intelligence, included sustained vowels and words with variations in length and articulation points. Twenty fluent speakers were invited, generating 148 audio files for training the model. The prototype had an accuracy of 90% in tests, superior to the FalaBrasil model(43-75%). In the second stage, a unigram model was used to analyze words in patients with speech apraxia, with 92.1% accuracy in recognition. The application was registered with the INPI and the Appraxi® trademark was granted.

Conclusions:

Appraxi® has proven effective in speech training, with acceptable accuracy, making it a useful tool in speech therapy in clinics, hospitals, or remote environments(minimum 4G internet connection). The technology is ready to assist in the recovery of patients with apraxia.