dc.description.abstract |
Human beings communicate with different systems in different ways but by considering the growth of technology speech communication becomes familiar and preferable to communicate with systems. In our country patient-physician ratio is not balanced this led to increase the burden of physicians in the health centers. In this domain an expert system needs to be implemented to solve existing problems of our country. Conversational systems which are developed for consulting about published low of Ethiopia and on hotel and restaurant domain text is used for conversation. The proposed study designs end-to-end speech based conversational AI for the domain of healthcare using under resourced language Amharic. To achieve our study, the required dataset is collected through interviewing pregnant women, physicians and nutrition professionals. Besides, we make an intensive literature review and analyze documents (manuals about pregnancy). A total of 560 text and audio data is prepared. We split the data into 80% for training and 20% for testing. The conversational system performs different tasks such as greeting, farewell, giving advice and suggestions, etc. Generally speaking, the proposed system achieved 92% accuracy. Thus, the proposed system shows an encouraging result with limited dataset. As future work developing Amharic speech recognizer that abeles to recognize different people’s with respect to age, gender and accent. A total of 560 text and audio data is prepared. We split the data into 80% for training and 20% for testing. The conversational system performs |
en_US |