| Summary: | 碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 97 === The Internet is full of many and complex web images, so it is hard to arrange these images to be searched. The one of solutions is automatic image annotation (AIA) and it can save a huge amount of manual effort. But the AIA is limited to the words that have been definition well.
In order to provide extra meaningful words for web images, in this study we use the latent semantic analysis (LSA) to analyze the web texts which are clustering together. And to explore extra word which is not including the web text itself. And the average precision is 50% and recall is 56%.
Because the LSA has to consume a lot of time to compute. In this study, we want a new web image can be annotated rapidly, so we use association analysis to train every clusters to get association rules. When a new web image is assigned to a appropriate cluster, the association rules which in this cluster would annotate this web image rapidly. The average precision is 79% and the recall is 57%.
|
|---|
