부산대학교 도서관

Online forum platforms have emerged as highly popular digital communities for seeking and sharing valuable information. Within these platforms, searches are typically constrained to keyword-based queries. However, the presence of noise in keywords, such as spelling errors or abbreviations, often impedes accurate information retrieval. The impact of this noise on search results poses challenges to both the effectiveness and efficiency of the retrieval process. This study introduces an ensemble similarity model that combines the Vector Space Model (VSM) with edit distance models to enhance thread analysis within a graduate discussion forum. The UMT Computer Science Forum System (UMTCSF) serves as an online platform for Computer Science students to engage in discussions encompassing questions, answers, and topic exploration. Operating in real-time through a web-based application, UMTCSF employs similarity analysis while effectively addressing keyword search noise. An ensemble similarity model demonstrates its potential and practicality in improving search result rankings. By adeptly combining different models and addressing keyword-related challenges, UMTCSF presents a valuable solution for facilitating efficient and accurate information retrieval within the realm of graduate forum discussions.