To evaluate the feedback to the patients during the naming training, 862 statements from 12 patients were used. These 862 statements were evaluated on the one hand by the ASR and on the other hand by 5 natural listeners. We were therefore able to make a total of 4310 comparisons between the automatic feedback and that from listeners. A first purely descriptive analysis showed that the average agreement between ASR and listener responses was 87.6%. The internally set goal of at least 80% agreement was clearly exceeded and confirms the impression that was created during the training.
Further, more detailed statistical evaluations will follow and will be published on this page in a timely manner.
With aphavox aphasia patients now have a web-based program for training support in self-training. Using end devices such as a tablet, aphavox.easy can be used wherever there is an online connection. The special highlight: A speech recognition mechanism conducts a dialogue with the patient and gives feedback on all inputs.
aphavox offers the option to practice naming in a direct form. The tasks for this are put together by the therapist. On the one hand, he can use the aphavox.easy settings menu, in which part of speech, word structure, word length and subject areas are defined. The program looks for appropriate exercise words according to these specifications and presents them in a random order.
As a second option, the words to be trained can also be selected directly and compiled into exercise units, so-called templates. The type, sequence and amount of support such as semantic cues, sentence additions, starting aids etc. are also determined by the therapist in the menu. In the event of errors, these can be activated automatically in a predetermined sequence or by the patient using the selection button.
Such classic or "confrontational" naming exercises can still be too difficult for many patients. However, in order to promote verbal language activation for them, the concept of decreasing assistance, vanishing cues, has been proven successful (eg Glisky, EL, 1992). aphavox.easy also offers this exercise option. First, the patient should recognize the target word from an image selection from up to four images. In the first step, he receives both the auditory specification and the written one. The written language is hidden at the next level of image selection. After this receptive "preliminary exercise", which is intended to prevent a purely empty repetition, the next step is a combined repetition and reading task. The image of the target word, the written word and the auditory specification are presented. The patient receives the request to repeat. At the next level, the font support is hidden, after that only the initial sound is given. Word additions, sentence additions follow at the end the pure naming task, whereby it is clear that this is often a delayed repetition. Even if the approach is still controversial (Abel et al., 2005), it seems to be a sensible alternative or addition to the classic naming with increasing help for some patients. It fits into the concept of error-free learning in aphasia therapy (Busch, T. & Heide, J., 2012).
The exercises that the patient completes with aphavox during the in-house training are individually tailored to him - according to the therapeutic requirements resulting from his diagnosis. This is possible because these exercises are put together by a therapist. For this, the therapist has access to databases with the exercise elements. With easy-to-use tools, he selects the right elements and puts them together for training, which he then assigns to his patients. At aphavox.easy, the patient only has access to the exercises intended for him. With this "extended arm", the therapist remains in control of the patient's own training.