PROBABILISTIC DIALOGUE MANAGER FOR MEDICAL DIAGNOSTIC

GEREZGIHER, WIDASSIE

dc.contributor.author	GEREZGIHER, WIDASSIE
dc.date.accessioned	2020-03-24T06:02:57Z
dc.date.available	2020-03-24T06:02:57Z
dc.date.issued	2020-03-24
dc.identifier.uri	http://hdl.handle.net/123456789/10784
dc.description.abstract	Medical dialogue system is the main stream application area of spoken dialogue system, which evolves dialogue management as main component with a role of decision making. The dialogue management stream in medical dialogue systems is still using the rule based and statistical approaches whereas spoken dialogue systems is using the hybrid approach, i.e. the probabilistic rules structure of modeling approach. In medical diagnosis physicians are in difficulty to diagnose an illness for closely related diseases, which are suspected to cause an illness. A mechanism to introduce principle of exclusiveness to these closely related diseases while diagnosing is indispensable. The objective is to develop probabilistic dialogue manager for medical diagnosis using Java with a user interface supporting both Amharic and English languages. The physician’s decision making behavior and patient’s feeling are modeled and uncertainties are handled. The PDMMD has used two models namely prior and posterior model. The prior model uses uniform probability distribution over the suspected diseases. The posterior model uses totality rule over probabilistic rules structure of modeling with Bayesian inference to select the system action using maximum function operation at run time. The PDMMD is evaluated with respect to literatures, and real physicians using questionnaires. A linear regression over the decision of physicians to the same scenario used by the system is used to evaluate the accuracy of the system. For closely related suspected diseases the system gives clear values to patient’s symptoms to represent for each suspected diseases, where the physicians show confusion in representing the symptoms clearly. Thus, the PDMMD outperforms in terms of simplicity, flexibility, dynamicity, uncertainty handling and accuracy for this domain and definitely helps to physicians. In addition to these, it also handles patient personal information and medication history. Lastly, it generates the report about the type of diagnostic method used, type of disease identified, and the condition of the patient and medication history.	en_US
dc.language.iso	en	en_US
dc.subject	Information Technology	en_US
dc.title	PROBABILISTIC DIALOGUE MANAGER FOR MEDICAL DIAGNOSTIC	en_US
dc.type	Thesis	en_US