Improving usability in access to information by a new artificial intelligence method

Information is one of the basic needs of each individual. Today, almost every individual spends most of their daily time on the internet to meet their various information needs or to exchange information. In this period that accessing information is much easier, it is thought that the point that needs to be focused more is "accessing useful information". The applications which we use the most about "accessing useful information" are search engines such as Google, Yandex, etc. These applications are considered under the name of "Information Retrieval Systems (IRS)" in the literature and they act as a bridge that brings the person who needs information and the source that contains the information together. Despite having various types, the common goal of IRS is to find out the information that will meet the needs. In this thesis, by developing an algorithm that can be integrated into any IRS infrastructure, we focused on providing an environment that can both update itself according to the various information needs of users and unify a currently used IRS infrastructure with more users. In this direction, for the purpose of achieving effective access to information, an artificial intelligence algorithm has been developed and has been integrated into site-in search tools operating within 170 websites belong to Kırklareli University (KLU). The aim of the study is to improve the information retrieval process independently of user language through real user behavior data and to evaluate the performance of the developed algorithm (Analogy) in this process. The Analogy algorithm developed within the scope of the thesis was trained with three months real user behavior data, and after that process, the contribution of the algorithm to the information retrieval process was analyzed in three different aspects (effectiveness, efficiency and satisfaction) in relation to the "Information Retrieval" and "Usability" literature. The results of the study showed that the Analogy algorithm made a significant contribution to the information retrieval process in terms of effectiveness and efficiency. The sign test results, which were carried out on the basis of the Reciprocal Rank metric, revealed the findings that indicate the success of the algorithm in terms of effectiveness (z=-2,421; p=0,015<0,05). Similar findings were revealed in the efficiency analysis based on three different indicators: "Response Time", "Time to First Click" and "Query Abandonment". In the analysis of Response Time based on the sign test, it was found that the algorithm responded more quickly to the users (z=-38,453; p=0,000<0,05). In the Time to First Click analysis, the sign test was also used, the findings indicating the success of the algorithm were revealed (z=-4,936; p=0,000<0,05). In the Query Abandonment indicator, which was analyzed using the Single Sample Chi-Square Test, the findings indicating the algorithm only lost a small number of users were revealed (χ2= 38,407; p=0,000<0,05). On the other hand, not enough data could be obtained to evaluate the algorithm in terms of satisfaction. It is thought that the Analogy algorithm, which takes action by learning from real user behavior regardless of user language, has the potential to be used not only in search engines but in many areas for effective and efficient access to useful information. Future researches are recommended to evaluate the usability of the algorithm in terms of satisfaction.