本文提出一個可以在具結構描述之物件中找出重複物件的方法。因為在比對找尋過程，物件結構與元素的缺值對物件相似與否會有影響。物件結構欄位不同，在比對時有不同的重要度；物件元素的缺值則會影響我們對物件的瞭解，缺值愈多，對物件的認識度愈低，可參考的資訊也愈少。所以本文針對物件結構採權重差異化，對元素缺值採可信度值處理，再以匯總的方式對比對之物件產生一匯總相似值，以判斷兩物件是否為重複物件。最後，以通訊錄匯總及人口普查實驗例子證實本方法可以有效的提升物件比對之準確率及召回率。 We propose a method for detecting the duplicates in structured objects. The structure of objects and the missing value of elements in objects are very important for detecting the duplicates. Different elements in a structured object have different importance in detecting duplicates. The missing value of elements influence us in understanding objects. The more of missing value, the less we understand the objects , and the less its reference value. So, we adopt the different weighting scheme of the structure of the object and compute the confidence value for missing values. Then, we summarize the similarity value and the confidence value of the two objects to decide if they are duplicates or not. Finally, we experiment with the synthetic addressbook and census datasets to test if our proposed method can improve the recall and precision in duplicate detection of objects.
