Search result clustering studies in Turkish

Information in the Internet increasing day by day and reaching desired information in this huge network is a major problem. In order to solve this problem, web search engines developed. Web search engines can index web sites to allow users search for the information they need on indexed data.If a large network such as internet is concerned, depends on the search query, a web search can return millions of web pages. Finding the most relevant information in this huge result set is a separate problem. Today?s search engines provide several ways to solve this problem. Generally; methods relies on sorting search results according to ranking with certain rules or clustering and grouping results developed.Search result clustering as well, is one of the methods to bring solution to this problem. It depends on applying various information extraction techniques to search results for clustering them by their content. The search results are displayed to users in clusters that contain descriptive labels. Thus, user selects one of the closest clusters to the information he/she need and he/she can access the information he need quicker.Many studies have been conducted for search result clustering. Suffix Tree Clustering (STC), one of the most widely used and fast algorithm in this field.Our aim in this thesis, performing search result clustering operations with Suffix Tree Clustering and another algorithm based on document similarity on Turkish web pages. Then, trying to improve these algorithms to get better results and measuring success of these algorithms.In our work, we used DBC which is based on document similarity calculation, classical STC, GSAK which is an imporved STC that we prepare and M-GSAK, which tries to improve GSAK results with using DBC results. On our tests, GSAK and M-GSAK get better results than classical SAK algorithm on Turkish web results. Depends on the F-Score calculations; GSAK scored 77% better performance than KSAK and M-GSAK scored 13% more than GSAK. Also we can say that; imporved STC techniques gets better results than document similarity calculation based DBC method. GSAK scored 9% better than DBC.