Архітектурне рішення для побудови інтелектуальних інтерфейсів користувача

Autor: Yu. O. Huchok, O. O. Nytrebych, Ye. V. Levus
Rok vydání: 2018
Předmět:
Zdroj: Науковий вісник НЛТУ України, Vol 28, Iss 8, Pp 155-160 (2018)
ISSN: 2519-2477
1994-7836
DOI: 10.15421/40280831
Popis: Traditional visual user interfaces are overloaded and not user friendly due to constant expanding of software functionality. The intelligent user interface (IUI) is an alternative to simplifying computer systems usage and solves a task of an interface personalization. Considering trends and shortcomings of existing solutions the relevant issue is a software concept development based on the "minimum interaction" approach which uses IUI. Either an Artificial Intelligence in the form of the remote Bayesian network or cloud technologies for Bayesian network usage in local software and for updating network knowledge obtained from individual local users or communication between a network and a local software with REST API are proposed as a solution for the issue mentioned. The purpose of the developed software is a creation of a flexible and simplified high-level system for the IUI aspects implementation in other software systems. This goal is implemented in the form of REST API service located on a remote server and designed for Bayesian networks construction. In terms of architecture the interface consists of three main parts: interaction management, knowledge base and authentication center. IUI parts are not located physically on a user device as they operate on a remote server instead. A knowledge base block is divided into two sub-blocks such as personal and general knowledge bases. The personal knowledge base contains all the information about the local user in a form of an object representation. The proposed method is implemented as "Personal Assistant" software type. The Bayesian network architecture with dynamic construction consists of the following three levels: the node level of the root "evidence" which includes the main network nodes that are time, geographical location, day of the week, etc.; the functional node level which displays the nodes of functions used by users; the function specification node level which includes all the additional parameters used in software function executions. The present method can be even more promising when not just bare statistics is taken into account, but other users' specifications at a deeper Artificial Intelligence level are considered as well as possible actions in terms of logic (not just in terms of statistics) are predicted.
Databáze: OpenAIRE
Externí odkaz: